{
 "dataset": {
  "archive": "dandi",
  "id": "000630",
  "label": "DANDI:000630"
 },
 "page": "datasets/dandi/000630/",
 "generated": "2026-11-05T07:41:23+00:00",
 "caution": "A failure to reproduce is not a finding that the paper is wrong. It may be a failure of the attempt.",
 "outcome_counts": {
  "reproduced": 9,
  "partially_reproduced": 0,
  "not_reproduced": 4,
  "blocked": 1
 },
 "outcome_counts_with_outside_data": {
  "reproduced": 1,
  "partially_reproduced": 1,
  "not_reproduced": 2,
  "blocked": 3
 },
 "outcome_counts_elsewhere": {
  "reproduced": 0,
  "partially_reproduced": 3,
  "not_reproduced": 0,
  "blocked": 15
 },
 "outcome_counts_links": {
  "reproduced": 0,
  "partially_reproduced": 2,
  "not_reproduced": 0,
  "blocked": 0
 },
 "attempts": [
  {
   "id": "000630-chartrand-2023-01",
   "dataset": {
    "archive": "dandi",
    "id": "000630",
    "version": "0.230915.2257",
    "doi": "10.48324/dandi.000630/0.230915.2257"
   },
   "key": "dandi:000630",
   "label": "DANDI:000630",
   "dataset_version": "0.230915.2257",
   "paper": "10.1126/science.adf0805",
   "paper_title": null,
   "date": "2026-10-31",
   "kind": "agent",
   "accountable": "CatalystNeuro",
   "agent": {
    "harness": "Claude Code",
    "model": "claude-opus-5-5",
    "protocol": "pilot-12"
   },
   "runner": {
    "platform": "GitHub Actions",
    "url": "https://github.com/catalystneuro/encore/actions/runs/41155476927"
   },
   "tier": "runs",
   "deprecated": "",
   "deprecated_on": null,
   "entry": "000630-chartrand-2023",
   "entry_title": "Human layer 1 interneuron electrophysiology (Chartrand et al. 2023)",
   "description": "Tests the electrophysiology findings of Chartrand et al. (2023) on dandiset 000630, which holds the current clamp recordings of the paper's 210 Seattle human L1 patch-seq cells. Features are extracted with IPFX from the long square sweeps and compared with the paper's per-cell table; cell types come from the paper's cell table, linked by the LIMS identifier. A second notebook compares the human cells with the paper's mouse L1 cells from dandiset 000020. A third notebook tests four morphology claims on the paper's deposited morphology features.",
   "notebooks": [
    {
     "path": "reproduce_chartrand_2023.ipynb",
     "title": "Reproducing the human electrophysiology findings of Chartrand et al. (2023) from dandiset 000630",
     "executed": "notebooks/000630-chartrand-2023/reproduce_chartrand_2023.html"
    },
    {
     "path": "mouse_comparison.ipynb",
     "title": "Human against mouse L1 cells, with dandiset 000020",
     "executed": "notebooks/000630-chartrand-2023/mouse_comparison.html"
    },
    {
     "path": "morphology_tables.ipynb",
     "title": "Morphology claims from the paper's deposited feature table",
     "executed": "notebooks/000630-chartrand-2023/morphology_tables.html"
    }
   ],
   "outcomes": [
    {
     "claim": "dandiset-cell-link",
     "statement": "The dandiset's files can be matched one to one to the cells of the paper's per-cell table, and the matched records agree on what both describe.",
     "source": "Data and materials availability; the code's data/human_l1_dataset_2023_02_06.csv and data/aibs_features_E.csv (github.com/AllenInstitute/patchseq_human_L1 at commit 7777857c1b89b19735e45fd9c9bd7493cc1d805c)",
     "quote": "",
     "statistics": [],
     "kind": "link",
     "inferred": "",
     "outcome": "partially_reproduced",
     "blocked_by": null,
     "notebook": "reproduce_chartrand_2023.ipynb",
     "notes": "All 210 files match one row each of the paper's cell table through the LIMS number (session_id equals the Ephys_Roi_Result number in the table's nwb_path); 204 files carry the table's cell name and 6 carry a time stamp in its place. Age agrees for all 208 cells with a known age: the files state ages as whole years times 365 days, and the one mismatch of the first run came from dividing by 365.25. The records do not all agree, so the partial level applies (at least 90% of files match and agree). One file (H19.03.321.12.04.01.04) states sex unknown where the table gives M, and the two files of donor 922967568 state an age of P0.0D where the table has none. Of the 44 features computed here, the 39 used in reading (a) agree with the paper's table within 1% for 53 to 98% of cells (Spearman 0.886 to 0.997); the five adaptation ratios not used agree for 4 to 37%. Where a feature at rheobase + 40 pA differs, the sweep chosen here is often another one than the paper's. For the initial firing rate at rheobase + 40 pA, 31 cells differ by more than 1% or have a value on one side only: 10 take another repeat of the same amplitude here, 6 another amplitude, 13 have no sweep here that gives the paper's value (5 of all cells have a different rheobase), and 2 have a value on one side only. In 16 cells no sweep passing QC lies 39 to 61 pA above rheobase and IPFX falls back to the spiking sweep closest to rheobase + 39 pA; 7 of them agree with the paper's value, and in 7 of the 9 that do not, the paper's value comes from a sweep in range that fails here only the Methods' 500 ms rms noise criterion. Without that criterion the 39 features agree with the paper's values for more cells in 38 of them (median share 0.933 against 0.895). The dandiset holds exactly the paper's primary set P (210 Seattle cells of the strict L1 set with long square features); the other 62 cells of the strict set, 50 Amsterdam and Szeged cells and 12 Seattle cells that failed feature extraction, are not in it.",
     "deviations": [
      "The match uses the LIMS number, as no file names its t-type or the table's specimen id."
     ],
     "assumed": false,
     "disagreed": true,
     "data": "with_labels",
     "where": null,
     "proposed": null
    },
    {
     "claim": "patchseq-counts",
     "statement": "The human L1 patch-seq dataset has 250 cells passing transcriptomic quality, 194 with electrophysiology passing quality control and 71 reconstructed; by subclass, 96 LAMP5, 35 MC4R, 33 PAX6, 15 L1 VIP and 15 other cells have electrophysiology.",
     "source": "Results, L1 patch-seq pipeline and transcriptomic references, paragraph 4; Table S2",
     "quote": "All cells were filtered for transcriptomic (n=250) and electrophysiological quality (n=194), and a subset of neurons (n=71) with sufficient cell labeling were imaged at high resolution",
     "statistics": [],
     "kind": "finding",
     "inferred": "",
     "outcome": "reproduced",
     "blocked_by": null,
     "notebook": "reproduce_chartrand_2023.ipynb",
     "notes": "194 cells of P in the ten L1 t-types; LAMP5 96, MC4R 35, PAX6 33, L1 VIP 15, other 15, as Table S2. The other 16 files of P are non-L1 t-types. The 250 and 71 counts need transcriptomes and reconstructions and are not tested.",
     "deviations": [
      "Cell labels (t-type, subclass, layer, region, donor) come from the paper's cell table through the link claim dandiset-cell-link."
     ],
     "assumed": false,
     "disagreed": false,
     "data": "with_labels",
     "where": null,
     "proposed": null
    },
    {
     "claim": "fig2c-subclass-ephys-differences",
     "statement": "Human L1 subclasses differ in electrophysiology, notably in sag and initial firing rate (Kruskal-Wallis, FDR-corrected p < 1e-7).",
     "source": "Results, Morpho-electric diversity in human L1, paragraph 1; Figure 2C",
     "quote": "Electrophysiologically, subclasses were distinguished by subthreshold properties such as sag ... as well as several suprathreshold properties including firing rate ... Electrophysiological and morphological features distinguishing L1 subclasses (KW test, FDR-corrected p<10\u22127 for electrophysiology",
     "statistics": [
      {
       "test": "Kruskal-Wallis test",
       "p": 1e-07,
       "p_relation": "<",
       "corrected": "FDR (Benjamini-Hochberg)",
       "n": "96 LAMP5, 35 MC4R, 33 PAX6 and 15 L1 VIP cells (Table S2)"
      }
     ],
     "kind": "finding",
     "inferred": "",
     "outcome": "reproduced",
     "blocked_by": null,
     "notebook": "reproduce_chartrand_2023.ipynb",
     "notes": "Kruskal-Wallis across the four core subclasses (179 cells, 92 features). Sag has FDR p = 6.6e-20 (epsilon2 0.57). The initial firing rate at rheobase + 40 pA (first_isi_inv_hero) has FDR p = 4.7e-7 on the features computed here with the Methods' sweep QC, above the claim's 1e-7, and 1.7e-8 on the paper's own values (reading (b)), below it. 56 features have FDR p < 0.05. The computed initial firing rate differs from the paper's value by more than 1%, or has a value on one side only, in 28 of the 179 cells: 10 take another repeat of the same amplitude, 6 another amplitude, 10 have no sweep here that gives the paper's value and 2 have a value on one side only. In 11 cells of the test no sweep passing QC lies 39 to 61 pA above rheobase and IPFX falls back to a lower sweep; 6 of them disagree with the paper, and in 4 of these the paper's value comes from a sweep in range that fails here only the Methods' 500 ms rms noise criterion. Extracted without that criterion (variant a'), the features agree with the paper's values for more cells in 38 of 39 features, and first_isi_inv_hero has FDR p = 3.9e-9 (sag 3.3e-20), so the claim holds in full. The outcome rests on (a'), an assumption stated under deviations; with the criterion as written the claim is partially reproduced. The outcome was partially_reproduced through round 2 and was changed to reproduced after verification round 3. With any one donor left out, the largest uncorrected p is 3.1e-19 for sag and 3.5e-5 for the initial firing rate, on reading (a); this check was not run on (a').",
     "deviations": [
      "Features follow the claims file's O1. Reading (a) uses the features computed here with IPFX 2.1.2 from the long square sweeps where they agree with the paper's aibs_features_E.csv within 1% for at least half of the cells (39 features, which agree within 1% for 53 to 98% of cells), and the paper's values for the others (custom features the paper did not release, ramp, short square and chirp features, and five adaptation ratios, which agree within 1% for only 4 to 37% of cells). Reading (b) uses the paper's values throughout. The 50% rule is an assumption of this attempt; the claims file's O1 does not say when a recomputed feature counts as computed. Two variants of reading (a) are shown, (a') with the features extracted without the Methods' 500 ms rms noise criterion, and (a, ratios) with the adaptation ratios computed here, for figs3a, fig6d and fig3g; neither changes an outcome except (a') for fig2c-subclass-ephys-differences, whose outcome rests on it.",
      "Assumed that the paper's sweep QC did not exclude the sweeps that fail here only the Methods' rms noise criterion of 0.5 mV over 500 ms (Methods, Quality control) as IPFX measures it, so the outcome is judged on variant (a'), extracted without that criterion. This departs from the Methods' QC as IPFX measures it. The paper's own per-cell values were computed from such sweeps (7 of the 9 fallback cells that disagree with the paper), and without the criterion 38 of the 39 recomputed features agree with the paper's values for more cells (median share within 1% 0.933 against 0.895). With the criterion as written (reading a) the claim is partially reproduced (first_isi_inv_hero FDR p 4.7e-7); on the paper's values (reading b) it holds (1.7e-8). How the paper measured this noise is asked of the authors (findings, methods_unavailable).",
      "Sweep QC follows the Methods as far as the files allow (leak within 100 pA, rms noise below 0.2 mV over 1.5 ms and 0.5 mV over 500 ms, IPFX's integrity check, and sweeps stored with zeros before the end of the stimulus left out); the Methods' criterion of a baseline within 2 mV of target is not applied, as the files do not state the target.",
      "Dunn's post hoc tests as the Methods and legend say; the code's Mann-Whitney post hoc is computed as a variant.",
      "Cell labels (t-type, subclass, layer, region, donor) come from the paper's cell table through the link claim dandiset-cell-link."
     ],
     "assumed": true,
     "disagreed": false,
     "data": "with_labels",
     "where": null,
     "proposed": null
    },
    {
     "claim": "fig2b-subclass-ephys-profiles",
     "statement": "LAMP5 cells show small sag and adaptation, PAX6 cells minimal sag and a high initial firing rate, while MC4R and L1 VIP cells show strong sag.",
     "source": "Results, Morpho-electric diversity in human L1, paragraph 2; Figure 2B and 2C",
     "quote": "Their electrophysiological phenotype was relatively undistinguished, with firing-rate adaptation and sag present but small ... along with minimal sag and high initial firing rate at the onset of response to depolarizing current injection ... MC4R cells had extremely compact ball-like axonal arbors, along with strong sag ... with a consistent stellate-like dendrite morphology and high sag",
     "statistics": [],
     "kind": "finding",
     "inferred": "",
     "outcome": "reproduced",
     "blocked_by": null,
     "notebook": "reproduce_chartrand_2023.ipynb",
     "notes": "Parts (i) and (ii) hold under both readings and in variant (a'), with median sag MC4R 0.336 and L1 VIP 0.277 above LAMP5 0.116 and PAX6 0.051, PAX6 lowest. Part (iii) as amended holds under both readings and in (a'), as PAX6's median initial firing rate is 49.3 Hz, above LAMP5 (31.8 to 32.1 Hz) and MC4R (28.8 to 30.5 Hz) with Dunn FDR p of 3.4e-5 or less. What the answer changed. The first author run, before the reader's answer and before verification, tested part (iii) as PAX6 having the highest median of the four and recorded not_reproduced, since L1 VIP is higher (54.6 Hz; Dunn FDR p 0.39 to 1 between PAX6 and L1 VIP). Asked whether that read the paper's \"high initial firing rate\" correctly, a reader amended part (iii) to PAX6 above LAMP5 and MC4R (questions/000630.yaml), because the paper does not compare PAX6 with L1 VIP on this feature. Re-run on the amended criterion (author-answers), the claim is reproduced, the outcome both verification rounds reviewed; the notebook still prints the earlier part (iii), which fails.",
     "deviations": [
      "Features follow the claims file's O1. Reading (a) uses the features computed here with IPFX 2.1.2 from the long square sweeps where they agree with the paper's aibs_features_E.csv within 1% for at least half of the cells (39 features, which agree within 1% for 53 to 98% of cells), and the paper's values for the others (custom features the paper did not release, ramp, short square and chirp features, and five adaptation ratios, which agree within 1% for only 4 to 37% of cells). Reading (b) uses the paper's values throughout. The 50% rule is an assumption of this attempt; the claims file's O1 does not say when a recomputed feature counts as computed. Two variants of reading (a) are shown, (a') with the features extracted without the Methods' 500 ms rms noise criterion, and (a, ratios) with the adaptation ratios computed here, for figs3a, fig6d and fig3g; neither changes an outcome except (a') for fig2c-subclass-ephys-differences, whose outcome rests on it.",
      "Cell labels (t-type, subclass, layer, region, donor) come from the paper's cell table through the link claim dandiset-cell-link."
     ],
     "assumed": false,
     "disagreed": false,
     "data": "with_labels",
     "where": null,
     "proposed": null
    },
    {
     "claim": "figs3a-sag-adaptation",
     "statement": "Spike adaptation properties are strongly and inversely related to sag across human L1 cells (R2 = 0.74).",
     "source": "Results, Morpho-electric diversity in human L1, paragraph 1; Figure S3A legend",
     "quote": "Spike adaptation properties showed a strong inverse relationship with sag across the dataset (Fig. S3A)",
     "statistics": [
      {
       "test": "linear regression",
       "statistic": "R2 = 0.74",
       "p": 1e-13,
       "p_relation": "<"
      }
     ],
     "kind": "finding",
     "inferred": "",
     "outcome": "not_reproduced",
     "blocked_by": null,
     "notebook": "reproduce_chartrand_2023.ipynb",
     "notes": "OLS of the upstroke adaptation ratio (third over first spike, the order that matches the paper's table) on sag with HC3 errors over 192 cells. In reading (a) the ratio is the paper's value (the ratio computed here agrees within 1% for 19% of cells and is not used) and sag is computed here: R2 = 0.334 (p = 1.3e-12). On the paper's values R2 = 0.359 (p = 3.3e-14), and in variant (a') 0.346, against 0.74 +/- 0.074. With both sag and the ratio computed here R2 is 0.204 (third over first, p = 1.6e-7) or 0.149 (first over third). All the paper's ratios are below 1 (0.38 to 0.97), so the upstroke falls from the first to the third spike, and the slope is negative (-0.61 in reading (a), -0.64 on the paper's values): cells with more sag have a larger fall, so more change, the opposite of the direction the criterion states. The paper says only inverse relationship, which a negative slope matches; whether the criterion misreads the paper is asked of a reader (questions/000630.yaml), and it cannot change the outcome, which fails on R2. On the medians of the ten t-types R2 is 0.28 to 0.31. The paper's own per-cell values with the code's regression do not give 0.74 either.",
     "deviations": [
      "The adaptation ratio is computed here on the lowest long square with at least 4 spikes (Methods); it agrees with the paper's value within 1% for 19% of cells (median difference 2.7%), so reading (a) uses the paper's ratio with sag computed here, and the regression on the ratio computed here, in both orders, is a variant.",
      "Features follow the claims file's O1. Reading (a) uses the features computed here with IPFX 2.1.2 from the long square sweeps where they agree with the paper's aibs_features_E.csv within 1% for at least half of the cells (39 features, which agree within 1% for 53 to 98% of cells), and the paper's values for the others (custom features the paper did not release, ramp, short square and chirp features, and five adaptation ratios, which agree within 1% for only 4 to 37% of cells). Reading (b) uses the paper's values throughout. The 50% rule is an assumption of this attempt; the claims file's O1 does not say when a recomputed feature counts as computed. Two variants of reading (a) are shown, (a') with the features extracted without the Methods' 500 ms rms noise criterion, and (a, ratios) with the adaptation ratios computed here, for figs3a, fig6d and fig3g; neither changes an outcome except (a') for fig2c-subclass-ephys-differences, whose outcome rests on it.",
      "Cell labels (t-type, subclass, layer, region, donor) come from the paper's cell table through the link claim dandiset-cell-link."
     ],
     "assumed": false,
     "disagreed": false,
     "data": "with_labels",
     "where": null,
     "proposed": null
    },
    {
     "claim": "fig2c-lamp5-gradient",
     "statement": "Within the LAMP5 subclass, sag and initial firing rate decrease from the NMBR to the DBP to the LCP2 t-type.",
     "source": "Results, Morpho-electric diversity in human L1, paragraph 2; Figure 2C",
     "quote": "Within the LAMP5 subclass, sag and initial firing rate decreased from the NMBR t-type to DBP to LCP2 t-types ... Spearman correlation FDR-corrected p<10\u22123",
     "statistics": [
      {
       "test": "Spearman correlation",
       "p": 0.001,
       "p_relation": "<",
       "corrected": "FDR"
      }
     ],
     "kind": "finding",
     "inferred": "",
     "outcome": "not_reproduced",
     "blocked_by": null,
     "notebook": "reproduce_chartrand_2023.ipynb",
     "notes": "Over the 96 LAMP5 cells of the strict set in the dandiset, both features have a positive Spearman rho with the code LCP2 = 0, DBP = 1, NMBR = 2, the direction the criterion asks for; the medians are highest in NMBR, but the initial firing rate is not monotone (DBP 18.6 Hz, LCP2 22.5 Hz). The initial firing rate meets the threshold (rho 0.46, FDR p = 2.6e-5; 3.0e-5 in variant a'; paper's values 2.5e-5). Sag does not (rho 0.28, FDR p = 0.022; 0.015 in a'; paper's values rho 0.36, FDR p = 0.0020; threshold 0.001). With the 14 Amsterdam and Szeged LAMP5 cells the code includes, taken from the paper's feature tables, sag has FDR p = 0.0039, still above 0.001. The test uses filter_dropouts=False, as the code's cell 59 calls run_correlations; with the function's default (True) sag, whose values are all below 1, is left out. The criterion names no partial level and sag fails under every reading, so the claim is not reproduced, by the rule applied to fig4b-adarb2-irregular-firing. The outcome was partially_reproduced through round 2 and was changed to not_reproduced after verification round 3.",
     "deviations": [
      "The dandiset lacks the Amsterdam and Szeged LAMP5 cells the code includes; the test runs on the Seattle cells, and the three-site test is a variant using the paper's per-cell features for the other sites.",
      "Features follow the claims file's O1. Reading (a) uses the features computed here with IPFX 2.1.2 from the long square sweeps where they agree with the paper's aibs_features_E.csv within 1% for at least half of the cells (39 features, which agree within 1% for 53 to 98% of cells), and the paper's values for the others (custom features the paper did not release, ramp, short square and chirp features, and five adaptation ratios, which agree within 1% for only 4 to 37% of cells). Reading (b) uses the paper's values throughout. The 50% rule is an assumption of this attempt; the claims file's O1 does not say when a recomputed feature counts as computed. Two variants of reading (a) are shown, (a') with the features extracted without the Methods' 500 ms rms noise criterion, and (a, ratios) with the adaptation ratios computed here, for figs3a, fig6d and fig3g; neither changes an outcome except (a') for fig2c-subclass-ephys-differences, whose outcome rests on it.",
      "Cell labels (t-type, subclass, layer, region, donor) come from the paper's cell table through the link claim dandiset-cell-link."
     ],
     "assumed": false,
     "disagreed": true,
     "data": "with_labels",
     "where": null,
     "proposed": null
    },
    {
     "claim": "figs6-donor-condition-ephys",
     "statement": "No electrophysiological feature of human L1 cells differs with donor medical condition, sex or age after FDR correction.",
     "source": "Results, Morpho-electric diversity in human L1, paragraph 3; Figure S6A and S6B legends",
     "quote": "AP width (upper right) was the only significant electrophysiology feature (MW p=0.03, NS after FDR correction)",
     "statistics": [
      {
       "about": "AP width, epileptic against non-epileptic donors, before correction",
       "test": "Mann-Whitney test",
       "p": 0.03
      },
      {
       "about": "sex, the only feature before correction",
       "test": "Mann-Whitney test",
       "p": 0.02
      }
     ],
     "kind": "finding",
     "inferred": "",
     "outcome": "reproduced",
     "blocked_by": null,
     "notebook": "reproduce_chartrand_2023.ipynb",
     "notes": "A claim of absence on the paper's tests (194 cells, 92 features). Medical condition, epileptic against not, smallest p 0.0027 (sag, r = 0.43), smallest FDR p 0.19; sex, smallest p 0.011 (rheobase, r = 0.22), smallest FDR p 0.98; age, 36 features after the dropout filter, smallest p 0.039 (AP height at rheobase + 40 pA, rho = -0.15), smallest FDR p 0.43. No feature has FDR p < 0.05 under either reading. Two limits weaken the absence of an effect. The public table gives a medical condition for only 86 of the 194 cells (65 epileptic, 21 not), so the condition test has less power than the paper's. The dropout filter the criterion takes from the code leaves 56 of the 92 features out of the age test, sag and three adaptation ratios among them; without the filter no feature reaches FDR p < 0.05 for age either (smallest FDR p 0.73, sag p = 0.11). Variant (a') gives the same (smallest FDR p 0.19, 0.99 and 0.43).",
     "deviations": [
      "The table's medical_conditions, donor_sex and donor_age stand in for the LIMS donor records the code reads, as the criterion says; 108 of 194 cells have no medical condition in the public table.",
      "Features follow the claims file's O1. Reading (a) uses the features computed here with IPFX 2.1.2 from the long square sweeps where they agree with the paper's aibs_features_E.csv within 1% for at least half of the cells (39 features, which agree within 1% for 53 to 98% of cells), and the paper's values for the others (custom features the paper did not release, ramp, short square and chirp features, and five adaptation ratios, which agree within 1% for only 4 to 37% of cells). Reading (b) uses the paper's values throughout. The 50% rule is an assumption of this attempt; the claims file's O1 does not say when a recomputed feature counts as computed. Two variants of reading (a) are shown, (a') with the features extracted without the Methods' 500 ms rms noise criterion, and (a, ratios) with the adaptation ratios computed here, for figs3a, fig6d and fig3g; neither changes an outcome except (a') for fig2c-subclass-ephys-differences, whose outcome rests on it.",
      "Cell labels (t-type, subclass, layer, region, donor) come from the paper's cell table through the link claim dandiset-cell-link."
     ],
     "assumed": false,
     "disagreed": false,
     "data": "with_labels",
     "where": null,
     "proposed": null
    },
    {
     "claim": "figs6c-region-ephys",
     "statement": "Some electrophysiological features of human L1 cells, including input resistance, vary with brain region; sparsely sampled regions either match MTG or have larger input resistance.",
     "source": "Results, Cross-species differences in L1, paragraph 2; Figure S6C",
     "quote": "We also explored dependence of L1 interneuron morpho-electric properties on brain region within our human data and found moderate effects on a set of features including dendrite extent and input resistance (Fig. S6C). For these features, more sparsely sampled regions either failed to differ from MTG cells or had smaller dendritic extent and larger input resistance",
     "statistics": [
      {
       "test": "Kruskal-Wallis test",
       "p": 0.05,
       "p_relation": "<",
       "corrected": "FDR"
      }
     ],
     "kind": "finding",
     "inferred": "",
     "outcome": "reproduced",
     "blocked_by": null,
     "notebook": "reproduce_chartrand_2023.ipynb",
     "notes": "Regions ITG (23 cells, 4 donors), MTG (115, 27), MFG (5, 3) and SFG (14, 5); 83 features, 6 with FDR p < 0.05. Input resistance (input_resistance_ss) has FDR p = 0.029; medians ITG 291, MTG 234, MFG 367, SFG 222 MOhm; no region is below MTG with Dunn FDR p < 0.05 (SFG against MTG p = 0.13; MFG above MTG, p = 0.0073). sag_peak_t, downstroke_rheo and avg_rate_max have FDR p 0.016 to 0.029. Variant (a') gives the same input resistance result (13 features with FDR p < 0.05 there).",
     "deviations": [
      "input_resistance_ss is not computed here and is the paper's value under both readings.",
      "Features follow the claims file's O1. Reading (a) uses the features computed here with IPFX 2.1.2 from the long square sweeps where they agree with the paper's aibs_features_E.csv within 1% for at least half of the cells (39 features, which agree within 1% for 53 to 98% of cells), and the paper's values for the others (custom features the paper did not release, ramp, short square and chirp features, and five adaptation ratios, which agree within 1% for only 4 to 37% of cells). Reading (b) uses the paper's values throughout. The 50% rule is an assumption of this attempt; the claims file's O1 does not say when a recomputed feature counts as computed. Two variants of reading (a) are shown, (a') with the features extracted without the Methods' 500 ms rms noise criterion, and (a, ratios) with the adaptation ratios computed here, for figs3a, fig6d and fig3g; neither changes an outcome except (a') for fig2c-subclass-ephys-differences, whose outcome rests on it.",
      "Cell labels (t-type, subclass, layer, region, donor) come from the paper's cell table through the link claim dandiset-cell-link."
     ],
     "assumed": false,
     "disagreed": false,
     "data": "with_labels",
     "where": null,
     "proposed": null
    },
    {
     "claim": "fig4b-mc4r-sag-irregularity",
     "statement": "MC4R (rosehip) cells are distinguished from the rest of human L1 by strong sag and by irregular firing (Mann-Whitney, FDR p < 1e-4).",
     "source": "Figure 4B legend; Results, Rosehip cells, paragraph 1",
     "quote": "Electrophysiological and morphological features distinguishing MC4R t-types (FDR-corrected MW test vs. rest of L1, p<10\u22124)",
     "statistics": [
      {
       "test": "Mann-Whitney test against rest of L1",
       "p": 0.0001,
       "p_relation": "<",
       "corrected": "FDR"
      }
     ],
     "kind": "finding",
     "inferred": "",
     "outcome": "reproduced",
     "blocked_by": null,
     "notebook": "reproduce_chartrand_2023.ipynb",
     "notes": "Reading (a) of the group (MC4R subclass, 35 of 210 Seattle strict-set cells, against the rest) gives sag median 0.336 against 0.118 (FDR p = 1.6e-18) and irregularity ratio 0.232 against 0.135 (FDR p = 1.2e-5), both higher and below 1e-4. Under reading (b) ADARB2 MC4R alone meets it (FDR p 3.1e-17 and 4.2e-6) and SST CHRNA4 alone does not (0.084 and 0.6). Variant (a') gives the same (FDR p 8.6e-19 and 1.2e-5).",
     "deviations": [
      "The irregularity ratio (ir_late_mean) is custom code the paper did not release; it is the paper's value under both readings.",
      "Features follow the claims file's O1. Reading (a) uses the features computed here with IPFX 2.1.2 from the long square sweeps where they agree with the paper's aibs_features_E.csv within 1% for at least half of the cells (39 features, which agree within 1% for 53 to 98% of cells), and the paper's values for the others (custom features the paper did not release, ramp, short square and chirp features, and five adaptation ratios, which agree within 1% for only 4 to 37% of cells). Reading (b) uses the paper's values throughout. The 50% rule is an assumption of this attempt; the claims file's O1 does not say when a recomputed feature counts as computed. Two variants of reading (a) are shown, (a') with the features extracted without the Methods' 500 ms rms noise criterion, and (a, ratios) with the adaptation ratios computed here, for figs3a, fig6d and fig3g; neither changes an outcome except (a') for fig2c-subclass-ephys-differences, whose outcome rests on it.",
      "Cell labels (t-type, subclass, layer, region, donor) come from the paper's cell table through the link claim dandiset-cell-link."
     ],
     "assumed": false,
     "disagreed": false,
     "data": "with_labels",
     "where": null,
     "proposed": null
    },
    {
     "claim": "fig4b-adarb2-irregular-firing",
     "statement": "Both rosehip t-types have strong sag, but only ADARB2 MC4R fires irregularly, with stronger and faster sag than SST CHRNA4.",
     "source": "Results, Rosehip cells, paragraph 1; Figure 4B",
     "quote": "Electrophysiologically, both t-types that comprise the MC4R subclass showed strong sag, but only the ADARB2 MC4R t-type showed the distinctive irregular firing (45) and stronger and faster sag",
     "statistics": [
      {
       "test": "Dunn's test post hoc to Kruskal-Wallis",
       "p": 0.05,
       "p_relation": "<",
       "corrected": "FDR"
      }
     ],
     "kind": "finding",
     "inferred": "",
     "outcome": "not_reproduced",
     "blocked_by": null,
     "notebook": "reproduce_chartrand_2023.ipynb",
     "notes": "(iii) holds, with median sag SST CHRNA4 0.234 and ADARB2 MC4R 0.374 above LAMP5 0.116 and PAX6 0.051. (iv) holds under both readings of faster, as ADARB2 MC4R has the higher sag and the shorter sag_peak_t (0.045 against 0.052 s) and sag_tau (0.050 against 0.090 s). (i) fails, as ADARB2 MC4R's median irregularity ratio is 0.421 against 0.153 in SST CHRNA4, but Dunn FDR p = 0.18 with 9 SST CHRNA4 cells. (ii) fails, as SST CHRNA4 (0.153) is above LAMP5 (0.114) with Dunn FDR p = 0.046. The criterion names no partial level, and the parts that fail are the claim's point that only ADARB2 MC4R fires irregularly; the direction of the irregularity difference holds, its significance does not. The first run recorded partially_reproduced; changed after verification round 1. Variant (a') gives the same.",
     "deviations": [
      "ir_late_mean, sag_peak_t and sag_tau are not computed here and are the paper's values under both readings.",
      "Features follow the claims file's O1. Reading (a) uses the features computed here with IPFX 2.1.2 from the long square sweeps where they agree with the paper's aibs_features_E.csv within 1% for at least half of the cells (39 features, which agree within 1% for 53 to 98% of cells), and the paper's values for the others (custom features the paper did not release, ramp, short square and chirp features, and five adaptation ratios, which agree within 1% for only 4 to 37% of cells). Reading (b) uses the paper's values throughout. The 50% rule is an assumption of this attempt; the claims file's O1 does not say when a recomputed feature counts as computed. Two variants of reading (a) are shown, (a') with the features extracted without the Methods' 500 ms rms noise criterion, and (a, ratios) with the adaptation ratios computed here, for figs3a, fig6d and fig3g; neither changes an outcome except (a') for fig2c-subclass-ephys-differences, whose outcome rests on it.",
      "Cell labels (t-type, subclass, layer, region, donor) come from the paper's cell table through the link claim dandiset-cell-link."
     ],
     "assumed": false,
     "disagreed": true,
     "data": "with_labels",
     "where": null,
     "proposed": null
    },
    {
     "claim": "fig5b-tnfaip8l3-ephys",
     "statement": "PAX6 TNFAIP8L3 cells are distinguished from all other human L1 cells by a high initial firing rate at rheobase and an after-depolarization following the last spike.",
     "source": "Results, Bursting PAX6 TNFAIP8L3 cells, paragraph 1; Figure 5B",
     "quote": "Both the initial firing rate at rheobase and the after-depolarization potential (ADP) following the final spike quantitatively distinguished PAX6 TNFAIP8L3 cells from all other L1 cells",
     "statistics": [
      {
       "test": "Mann-Whitney test against rest of L1",
       "p": 0.01,
       "p_relation": "<",
       "corrected": "FDR",
       "n": "2 PAX6 TNFAIP8L3 cells with electrophysiology (Table S2) against the rest of L1"
      }
     ],
     "kind": "finding",
     "inferred": "",
     "outcome": "reproduced",
     "blocked_by": null,
     "notebook": "reproduce_chartrand_2023.ipynb",
     "notes": "Two PAX6 TNFAIP8L3 cells against 208. The initial firing rate at rheobase (computed here, equal to the paper's values) is 136 and 123 Hz against a median of 6.5 Hz (FDR p = 0.0081); the ADP slope (postap_slope_rheo, the paper's value) is 2.4 and 2.85 V/s against -0.43 (FDR p = 0.0051). Both are higher and below 0.01. The adp_deltav variant has FDR p = 0.07. Variant (a') gives the same (FDR p 0.0079 and 0.0051).",
     "deviations": [
      "postap_slope_rheo and adp_v_last_rheo are custom features the paper did not release; they are the paper's values.",
      "Features follow the claims file's O1. Reading (a) uses the features computed here with IPFX 2.1.2 from the long square sweeps where they agree with the paper's aibs_features_E.csv within 1% for at least half of the cells (39 features, which agree within 1% for 53 to 98% of cells), and the paper's values for the others (custom features the paper did not release, ramp, short square and chirp features, and five adaptation ratios, which agree within 1% for only 4 to 37% of cells). Reading (b) uses the paper's values throughout. The 50% rule is an assumption of this attempt; the claims file's O1 does not say when a recomputed feature counts as computed. Two variants of reading (a) are shown, (a') with the features extracted without the Methods' 500 ms rms noise criterion, and (a, ratios) with the adaptation ratios computed here, for figs3a, fig6d and fig3g; neither changes an outcome except (a') for fig2c-subclass-ephys-differences, whose outcome rests on it.",
      "Cell labels (t-type, subclass, layer, region, donor) come from the paper's cell table through the link claim dandiset-cell-link."
     ],
     "assumed": false,
     "disagreed": false,
     "data": "with_labels",
     "where": null,
     "proposed": null
    },
    {
     "claim": "fig6a-ephys-dprime-subclass-structure",
     "statement": "Pairwise electrophysiological distinctness (d') of human L1 t-types is smaller within subclasses than between them; the only within-subclass pair with high d' is LAMP5 NMBR against LAMP5 LCP2.",
     "source": "Results, Cross-modality relationships of L1 subclasses and t-types, paragraph 1; Figure 6A",
     "quote": "The resulting t-type similarity matrices (Fig. 6A) showed comparable subclass structure in both electrophysiology and transcriptomics, with smaller d\u2019 values within subclass blocks and higher values outside ... The single within-subclass pair with a high d\u2019 was LAMP5 NMBR and LAMP5 LCP2",
     "statistics": [],
     "kind": "finding",
     "inferred": "",
     "outcome": "not_reproduced",
     "blocked_by": null,
     "notebook": "reproduce_chartrand_2023.ipynb",
     "notes": "Nine t-types with at least 5 cells (the legend's rule of at least 4 keeps the same nine), 81 features. Mean within-subclass d' is below the mean between-subclass d' in all nine runs (features computed here 1.24 to 1.31 against 2.48 to 2.54, variant a' 1.28 to 1.30 against 2.52 to 2.56, paper's values 1.12 to 1.23 against 2.39 to 2.49, seeds 0 to 2). The criterion asks that NMBR against LCP2 be the only high within-subclass pair under both of its readings of high. With high as d' >= 2.2 it is so on the features computed here and in (a') for every seed (2.39 for seed 0) and on the paper's values for seed 0 only (no within-subclass pair reaches 2.2 for seeds 1 and 2). With high as above the median between-subclass d' (2.40 to 2.62) no within-subclass pair is high. The subclass structure holds; the single high pair holds only under the first reading. The first run adopted the 2.2 reading as an assumption and recorded reproduced; changed to partially_reproduced after verification round 1, as the criterion asks for both readings. The criterion names no partial level, so the claim is not reproduced; changed to not_reproduced after verification round 3.",
     "deviations": [
      "Random forest seeds 0, 1 and 2 are a standard choice; the code sets none.",
      "Features follow the claims file's O1. Reading (a) uses the features computed here with IPFX 2.1.2 from the long square sweeps where they agree with the paper's aibs_features_E.csv within 1% for at least half of the cells (39 features, which agree within 1% for 53 to 98% of cells), and the paper's values for the others (custom features the paper did not release, ramp, short square and chirp features, and five adaptation ratios, which agree within 1% for only 4 to 37% of cells). Reading (b) uses the paper's values throughout. The 50% rule is an assumption of this attempt; the claims file's O1 does not say when a recomputed feature counts as computed. Two variants of reading (a) are shown, (a') with the features extracted without the Methods' 500 ms rms noise criterion, and (a, ratios) with the adaptation ratios computed here, for figs3a, fig6d and fig3g; neither changes an outcome except (a') for fig2c-subclass-ephys-differences, whose outcome rests on it.",
      "Cell labels (t-type, subclass, layer, region, donor) come from the paper's cell table through the link claim dandiset-cell-link."
     ],
     "assumed": false,
     "disagreed": true,
     "data": "with_labels",
     "where": null,
     "proposed": null
    },
    {
     "claim": "fig6d-gmm-subclass-f1",
     "statement": "Two-cluster Gaussian mixtures on the best pair of electrophysiological features recover each subclass with F1 scores of 0.81 (LAMP5), 0.69 (MC4R), 0.89 (PAX6) and 0.5 (all others).",
     "source": "Results, Cross-modality relationships of L1 subclasses and t-types, paragraph 2; Figure 6D legend",
     "quote": "F1 scores ... LAMP5 0.81, MC4R 0.69, PAX6 0.89, all others 0.5",
     "statistics": [],
     "kind": "finding",
     "inferred": "",
     "outcome": "reproduced",
     "blocked_by": null,
     "notebook": "reproduce_chartrand_2023.ipynb",
     "notes": "177 Seattle cells with soma in L1 (the depth table's layer, as clustering_classification.ipynb cell 16), 83 features, 3486 one- and two-feature subspaces per group. On the features computed here, Methods preprocessing (O3 a) gives best F1 LAMP5 0.785, MC4R 0.709, PAX6 0.896, all others 0.490; the code's preprocessing (O3 b) gives 0.806, 0.703, 0.892 and 0.526. On the paper's values alone the two give LAMP5 0.802 to 0.806, MC4R 0.693 to 0.712, PAX6 0.892 to 0.896 and all others 0.495 to 0.506. All are within tolerance; the closest to a bound is all others at 0.490 (floor 0.45). With the layer filled from the LIMS layer (189 cells, code preprocessing) the values are 0.840, 0.707, 0.879 and 0.494; in variant (a') 0.806, 0.703, 0.892 and 0.519; with the adaptation ratios computed here 0.825, 0.694, 0.892 and 0.511. Several best features are the paper's values (sag_tau, sag_peak_t, ramp and short square features).",
     "deviations": [
      "Features follow the claims file's O1. Reading (a) uses the features computed here with IPFX 2.1.2 from the long square sweeps where they agree with the paper's aibs_features_E.csv within 1% for at least half of the cells (39 features, which agree within 1% for 53 to 98% of cells), and the paper's values for the others (custom features the paper did not release, ramp, short square and chirp features, and five adaptation ratios, which agree within 1% for only 4 to 37% of cells). Reading (b) uses the paper's values throughout. The 50% rule is an assumption of this attempt; the claims file's O1 does not say when a recomputed feature counts as computed. Two variants of reading (a) are shown, (a') with the features extracted without the Methods' 500 ms rms noise criterion, and (a, ratios) with the adaptation ratios computed here, for figs3a, fig6d and fig3g; neither changes an outcome except (a') for fig2c-subclass-ephys-differences, whose outcome rests on it.",
      "Cell labels (t-type, subclass, layer, region, donor) come from the paper's cell table through the link claim dandiset-cell-link."
     ],
     "assumed": false,
     "disagreed": false,
     "data": "with_labels",
     "where": null,
     "proposed": null
    },
    {
     "claim": "figs8a-classifier-cv",
     "statement": "A multi-class classifier assigns human L1 subclasses from electrophysiology with 82% cross-validated accuracy, balanced across subclasses.",
     "source": "Results, Cross-modality relationships of L1 subclasses and t-types, paragraph 2; Figure S8A",
     "quote": "A multi-class classifier evaluated by cross-validation on the primary dataset had 82% accuracy balanced across subclasses",
     "statistics": [],
     "kind": "finding",
     "inferred": "",
     "outcome": "reproduced",
     "blocked_by": null,
     "notebook": "reproduce_chartrand_2023.ipynb",
     "notes": "164 Seattle cells of the four core subclasses with soma in L1 (depth table layer, as the code's cell 26). Accuracy with features common to all sites (72) is 83.5% with the Methods' preprocessing and 84.6% with the code's; with the 35 site-aligned features 79.7% and 78.8%. On the paper's values (reading b) and with the layer filled from the LIMS layer (176 cells) accuracy is 79.6 to 84.2%, and in variant (a') 79.4 to 85.0%. All are within 82 +/- 8.2. Balanced accuracy is 74.5 to 79.0%. The largest class is 52.4% of cells (54.0% with the filled layer) and balanced chance is 25%; the paper states no chance level.",
     "deviations": [
      "Choosing features by site uses the Amsterdam and Szeged feature tables of the paper's code (tamas_features_E.csv, mansvelder_features_E.csv); those cells are not in the dandiset and are used only for this choice.",
      "Repeated stratified 5-fold cross-validation with 10 repeats and seed 0 (scikit-learn defaults; the seed is a standard choice).",
      "Features follow the claims file's O1. Reading (a) uses the features computed here with IPFX 2.1.2 from the long square sweeps where they agree with the paper's aibs_features_E.csv within 1% for at least half of the cells (39 features, which agree within 1% for 53 to 98% of cells), and the paper's values for the others (custom features the paper did not release, ramp, short square and chirp features, and five adaptation ratios, which agree within 1% for only 4 to 37% of cells). Reading (b) uses the paper's values throughout. The 50% rule is an assumption of this attempt; the claims file's O1 does not say when a recomputed feature counts as computed. Two variants of reading (a) are shown, (a') with the features extracted without the Methods' 500 ms rms noise criterion, and (a, ratios) with the adaptation ratios computed here, for figs3a, fig6d and fig3g; neither changes an outcome except (a') for fig2c-subclass-ephys-differences, whose outcome rests on it.",
      "Cell labels (t-type, subclass, layer, region, donor) come from the paper's cell table through the link claim dandiset-cell-link."
     ],
     "assumed": false,
     "disagreed": false,
     "data": "with_labels",
     "where": null,
     "proposed": null
    },
    {
     "claim": "figs8b-classifier-holdout",
     "statement": "Trained on the primary dataset with site-dependent features excluded, the classifier reaches 81% accuracy on held-out cells recorded at other sites.",
     "source": "Results, Cross-modality relationships of L1 subclasses and t-types, paragraph 2; Figure S8B",
     "quote": "After excluding features for which the distributions strongly differed from the primary dataset, we found comparable classification performance (81%, Fig. S8B)",
     "statistics": [],
     "kind": "finding",
     "inferred": "",
     "outcome": "blocked",
     "blocked_by": "data_missing",
     "notebook": "reproduce_chartrand_2023.ipynb",
     "notes": "The 26 held-out cells (17 Szeged, 9 Amsterdam; strict set, core subclasses, soma in L1) are not in the dandiset (it holds only the 210 Seattle cells of P), and the paper names no other deposit of their recordings; the criterion blocks the claim as data_missing in that case. Two DANDI datasets published in 2026 were checked from their metadata. DANDI 000228 (version 0.260829.2040) holds 91 Szeged files from 32 donors, including the donors of 15 of the 17 held-out Szeged cells, but its file paths carry no session id, so they could be linked to cells only by matching features. DANDI 000337 (version 0.260917.2218) holds 21 Amsterdam files; none has the LIMS session id of a held-out Amsterdam cell (5 of the 9 have one in the table), and 8 carry the session id PLACEHOLDER in their paths. Whether the claims file should name 000228 as an input is asked of a reader (questions/000630.yaml).",
     "deviations": [],
     "assumed": false,
     "disagreed": false,
     "data": "with_labels",
     "where": null,
     "proposed": null
    },
    {
     "claim": "fig3d-ephys-species",
     "statement": "Mouse L1 cells have less sag, lower spike amplitude and higher rheobase than human, across subclasses and within at least 2 of the 4 matched subclasses.",
     "source": "Results, Cross-species differences in L1, paragraph 2; Figure 3D right",
     "quote": "Mouse cells had less sag, lower amplitude spikes, and higher rheobase across types and within at least 2 of 4 matched subclasses",
     "statistics": [
      {
       "test": "2-way ANOVA on ranks, species effect",
       "p": 1e-38,
       "p_relation": "<",
       "corrected": "FDR"
      }
     ],
     "kind": "finding",
     "inferred": "",
     "outcome": "partially_reproduced",
     "blocked_by": null,
     "notebook": "mouse_comparison.ipynb",
     "notes": "On the cells the claims file names, human cells of P with soma in L1 (176, the layer filled from the LIMS layer as l1_cross_species.ipynb cell 5 does) and mouse VISp L1 cells (150), 80 features. Directions hold for all three features (mouse sag 0.048 against 0.136, AP height at the ramp 54 against 76 mV, rheobase 140 against 30 pA), with species FDR p < 0.05, and for each feature with an interaction at least 2 of the 4 subclasses differ at p < 0.05 in that direction (LAMP5 and MC4R for all three, L1 VIP for sag and rheobase). AP height has species FDR p 3.4e-33 under O2 reading (a) and 3.3e-34 under (b), and sag 9.5e-36 under (a), so not all three are below 1e-38 under either reading; the readings agree on partially reproduced. The one mouse PAX6 cell in VISp leaves PAX6 out of the ANOVA. The first run used every visual area (VIS, 185 mouse cells), as the code does, where the readings disagree (AP height 2.6e-38 under (a), 1.4e-43 under (b)) and the claim was blocked as methods_underspecified; changed after verification round 1 to the claims file's VISp. With the 22 strict mouse cells missing from DANDI 000020 added from the paper's table (168 VISp cells) the readings disagree as well (sag 2.7e-36 under (a), all three below 1e-38 under (b)), so the missing cells lower the power of the test. With the human layer not filled (164 cells, as the code's other notebooks select) and VISp, AP height has species FDR p 9.3e-32 and 3.7e-33, also above 1e-38 under both readings. Variant (a') gives the same on VISp (AP height 3.4e-33 and 3.3e-34, sag 3.1e-36 under O2 a).",
     "deviations": [
      "AP height at the ramp (peak_deltav_ramp) is not computed here and is the paper's value for both species.",
      "Features follow the claims file's O1. Reading (a) uses the features computed here with IPFX 2.1.2 from the long square sweeps where they agree with the paper's aibs_features_E.csv within 1% for at least half of the cells (38 features computed for both species; 39 for the human cells alone, which agree within 1% for 53 to 98% of cells), and the paper's values for the others (custom features the paper did not release, ramp, short square and chirp features, and five adaptation ratios, which agree within 1% for only 4 to 37% of cells). Reading (b) uses the paper's values throughout. The 50% rule is an assumption of this attempt; the claims file's O1 does not say when a recomputed feature counts as computed. Two variants of reading (a) are shown, (a') with the features extracted without the Methods' 500 ms rms noise criterion, and (a, ratios) with the adaptation ratios computed here, for figs3a, fig6d and fig3g; neither changes an outcome except (a') for fig2c-subclass-ephys-differences, whose outcome rests on it.",
      "Mouse cells from DANDI 000020 version 0.210913.1639 with the paper's mouse labels; 250 of 272 strict mouse L1 cells have a file there.",
      "Cell labels (t-type, subclass, layer, region, donor) come from the paper's cell table through the link claim dandiset-cell-link.",
      "Mouse VISp cells as the claims file says; the code's mouse set takes every visual area (VIS), shown as a variant.",
      "A subclass with fewer than 2 cells of either species with a value is left out of that feature's ANOVA, as the HC3 errors are undefined with one cell; with VISp this leaves out PAX6 (1 mouse cell) for every feature."
     ],
     "assumed": false,
     "disagreed": false,
     "data": "with_inputs",
     "where": null,
     "proposed": null
    },
    {
     "claim": "fig3e-tea-validation",
     "statement": "The human versus mouse electrophysiological differences found against VISp are replicated against mouse TEa, so they are not regional.",
     "source": "Results, Cross-species differences in L1, paragraph 2; Figure 3E",
     "quote": "We replicated these findings in a comparison between MTG and a smaller TEa dataset, verifying that cross species differences were not due to regional differences between MTG and VISp",
     "statistics": [
      {
       "test": "Mann-Whitney test",
       "p": 0.05,
       "p_relation": "<",
       "corrected": "FDR"
      }
     ],
     "kind": "finding",
     "inferred": "",
     "outcome": "blocked",
     "blocked_by": "data_missing",
     "notebook": "mouse_comparison.ipynb",
     "notes": "The mouse TEa comparison needs the Amsterdam TEa recordings (the code reads mansvelder_tea_2022_09_16.csv from a local project folder, not in its repository) and the Allen TEa cells. The paper names no deposit for either; the 15 TEa cells of the paper's mouse table all have a LIMS id, but none has a file in DANDI 000020 version 0.210913.1639, and the code's repository holds no TEa features.",
     "deviations": [],
     "assumed": false,
     "disagreed": false,
     "data": "with_inputs",
     "where": null,
     "proposed": null
    },
    {
     "claim": "fig3f-a-type-conductance",
     "statement": "In nucleated patches, mouse L1 neurons (only LAMP5 cells were sampled in mouse) have much higher A-type K+ conductance than human, with comparable D-type conductance.",
     "source": "Results, Cross-species differences in L1, paragraph 3; Figure 3F",
     "quote": "Compared with human neurons, mouse neurons showed much higher A-type K+ conductance but comparable slow inactivating (D-type) conductance",
     "statistics": [],
     "kind": "finding",
     "inferred": "",
     "outcome": "blocked",
     "blocked_by": "data_missing",
     "notebook": "mouse_comparison.ipynb",
     "notes": "Nucleated-patch voltage clamp sweeps (stimulus codes starting with Nuc) are in 38 of the 210 human files and in 25 of the 250 strict mouse L1 files of DANDI 000020 (15 of them LAMP5 cells), so IKA and IKD could be computed for both species. The conductance densities the claim compares need each patch's surface area, measured from images that are not public, so the claim stays blocked as data_missing. The conductances themselves were not computed here.",
     "deviations": [],
     "assumed": false,
     "disagreed": false,
     "data": "with_inputs",
     "where": null,
     "proposed": null
    },
    {
     "claim": "fig3g-variability-dimensions",
     "statement": "Human and mouse L1 subclasses have a similar amount of morpho-electric variability but along different features; sag and spike shape adaptation vary in human but little in mouse, while ISI adaptation and AHP vary in mouse.",
     "source": "Results, Cross-species differences in L1, paragraph 4; Figure 3G",
     "quote": "we found that the two species had a similar amount of variability (number of significantly different features and their effect size) but varied along different sets of features (Fig. 3G). The most distinct features in human, like sag and spike shape adaptation, showed little variability in mouse, and unlike in human, mouse subclasses varied physiologically in ISI adaptation and spike after-hyperpolarization (AHP) properties",
     "statistics": [],
     "kind": "finding",
     "inferred": "",
     "outcome": "reproduced",
     "blocked_by": null,
     "notebook": "mouse_comparison.ipynb",
     "notes": "Human 179 cells (92 features), mouse 205 VISp cells (82 features), as the claims file says. Sag and the upstroke adaptation ratio vary across human subclasses (epsilon2 0.57 and 0.46, FDR p < 1e-4) and less in mouse (0.22 and 0.17). In mouse the ISI adaptation features isi_adapt_ratio, adapt_mean and adapt_hero (epsilon2 0.38 to 0.41) and the AHP features fast_trough_deltav_rheo, fast_trough_deltav_hero and trough_slowdeltav_hero (0.31 to 0.40) have FDR p < 1e-4 and larger epsilon2 than in human. Both feature readings agree, and the same holds in variant (a'), with the adaptation ratios computed here (upstroke adaptation ratio epsilon2 0.43 in human against 0.18 in mouse) and with every visual area (VIS, 250 cells), which the first run used. The amount, outside the outcome, is 56 FDR-significant features in human and 73 in mouse VISp (77 in VIS), mean epsilon2 0.17 and 0.22.",
     "deviations": [
      "Features follow the claims file's O1. Reading (a) uses the features computed here with IPFX 2.1.2 from the long square sweeps where they agree with the paper's aibs_features_E.csv within 1% for at least half of the cells (38 features computed for both species; 39 for the human cells alone, which agree within 1% for 53 to 98% of cells), and the paper's values for the others (custom features the paper did not release, ramp, short square and chirp features, and five adaptation ratios, which agree within 1% for only 4 to 37% of cells). Reading (b) uses the paper's values throughout. The 50% rule is an assumption of this attempt; the claims file's O1 does not say when a recomputed feature counts as computed. Two variants of reading (a) are shown, (a') with the features extracted without the Methods' 500 ms rms noise criterion, and (a, ratios) with the adaptation ratios computed here, for figs3a, fig6d and fig3g; neither changes an outcome except (a') for fig2c-subclass-ephys-differences, whose outcome rests on it.",
      "Mouse cells from DANDI 000020 version 0.210913.1639; 250 of 272 strict mouse L1 cells have a file there, 205 of them in VISp.",
      "Cell labels (t-type, subclass, layer, region, donor) come from the paper's cell table through the link claim dandiset-cell-link."
     ],
     "assumed": false,
     "disagreed": false,
     "data": "with_inputs",
     "where": null,
     "proposed": null
    },
    {
     "claim": "fig4d-mouse-no-rosehip-ephys",
     "statement": "The mouse MC4R subclass has moderate sag but no irregular spiking; irregular spiking occurs only in a subset of mouse LAMP5 cells (mainly Lamp5 Ntn1 Npy2r) without other rosehip features.",
     "source": "Results, Rosehip cells, paragraph 2; Figure 4D",
     "quote": "the homologous mouse MC4R subclass had moderate sag but no irregular spiking (Fig. 4D). Irregular spiking resembling the human ADARB2 MC4R rosehip t-type was present only in a subset of LAMP5 cells (primarily Lamp5 Ntn1 Npy2r) that did not have other rosehip-like features",
     "statistics": [],
     "kind": "finding",
     "inferred": "",
     "outcome": "not_reproduced",
     "blocked_by": null,
     "notebook": "mouse_comparison.ipynb",
     "notes": "235 mouse VIS cells passing QC. MC4R median sag 0.141 to 0.147 is above LAMP5's 0.041 to 0.042 (Dunn FDR p about 1e-9). 4 of 18 MC4R cells have an irregularity ratio above 0.4, so reading (a), none, fails and reading (b), a minority, holds. Of the 30 cells above 0.4, 15 are LAMP5, 11 L1 VIP and 4 MC4R (not mostly LAMP5), and the largest t-type among them is Lamp5 Plch2 Dock5 (8) before Lamp5 Ntn1 Npy2r (6). These two parts fail under both readings, and the criterion names no partial level. With all 272 strict cells from the paper's table, 18 of 34 irregular cells are LAMP5, a majority, but the largest t-type is still Lamp5 Plch2 Dock5 (11), so the outcome is the same; cuts of 0.3 and 0.5 give the same order of t-types or put Vip Col15a1 Pde1a first. Variant (a') gives the same counts. The moderate sag holds. The first run recorded partially_reproduced; changed after verification round 1.",
     "deviations": [
      "ir_late_mean is the paper's value (custom code not released).",
      "Features follow the claims file's O1. Reading (a) uses the features computed here with IPFX 2.1.2 from the long square sweeps where they agree with the paper's aibs_features_E.csv within 1% for at least half of the cells (38 features computed for both species; 39 for the human cells alone, which agree within 1% for 53 to 98% of cells), and the paper's values for the others (custom features the paper did not release, ramp, short square and chirp features, and five adaptation ratios, which agree within 1% for only 4 to 37% of cells). Reading (b) uses the paper's values throughout. The 50% rule is an assumption of this attempt; the claims file's O1 does not say when a recomputed feature counts as computed. Two variants of reading (a) are shown, (a') with the features extracted without the Methods' 500 ms rms noise criterion, and (a, ratios) with the adaptation ratios computed here, for figs3a, fig6d and fig3g; neither changes an outcome except (a') for fig2c-subclass-ephys-differences, whose outcome rests on it.",
      "22 of the paper's strict mouse L1 cells (20 in VISp layer 1) have no file in DANDI 000020.",
      "Cell labels (t-type, subclass, layer, region, donor) come from the paper's cell table through the link claim dandiset-cell-link."
     ],
     "assumed": false,
     "disagreed": true,
     "data": "with_inputs",
     "where": null,
     "proposed": null
    },
    {
     "claim": "fig5d-mouse-pax6-doublets",
     "statement": "Mouse PAX6 subclass cells (Lamp5 Krt73) fire doublets at stimulus onset and show a depolarizing hump just below rheobase, marking them as alpha7 cells.",
     "source": "Results, Bursting PAX6 TNFAIP8L3 cells, paragraph 2; Figure 5D",
     "quote": "In mouse L1, the homologous PAX6 subclass was extremely rare, comprising only a few cells in the Lamp5 Krt73 t-type. These cells tended to fire in doublets at stimulus onset ... Mouse doublet-firing cells also showed a depolarizing \u2018hump\u2019 for current injection just below rheobase",
     "statistics": [],
     "kind": "finding",
     "inferred": "",
     "outcome": "not_reproduced",
     "blocked_by": null,
     "notebook": "mouse_comparison.ipynb",
     "notes": "Three Lamp5 Krt73 cells are in DANDI 000020 (the claims file lists 4; the fourth, specimen 710346584 in VISp1, has no file and no value for either feature in the paper's table), two with a value at rheobase + 40 pA. Their depolarizing sag ratios (0.257, 0.273, 0.304) exceed the medians of every other mouse subclass (0.035 to 0.132). Their median initial firing rate, 97.7 Hz (58 and 137 Hz), exceeds LAMP5 (20 Hz) and MC4R (26 Hz) but not L1 VIP (171 to 177 Hz), under both readings and with all 272 strict cells. Variant (a') gives the same medians. The hump holds; the doublet part of the criterion does not, and the criterion names no partial level. The paper says only that these cells tended to fire in doublets at stimulus onset and does not compare them with L1 VIP; whether the criterion's comparison with every other subclass misreads the paper is asked of a reader (questions/000630.yaml). Under the narrower reading, above LAMP5 and MC4R, both parts hold. The first run recorded partially_reproduced; changed after verification round 1.",
     "deviations": [
      "sag_depol is the paper's value (custom code not released).",
      "Features follow the claims file's O1. Reading (a) uses the features computed here with IPFX 2.1.2 from the long square sweeps where they agree with the paper's aibs_features_E.csv within 1% for at least half of the cells (38 features computed for both species; 39 for the human cells alone, which agree within 1% for 53 to 98% of cells), and the paper's values for the others (custom features the paper did not release, ramp, short square and chirp features, and five adaptation ratios, which agree within 1% for only 4 to 37% of cells). Reading (b) uses the paper's values throughout. The 50% rule is an assumption of this attempt; the claims file's O1 does not say when a recomputed feature counts as computed. Two variants of reading (a) are shown, (a') with the features extracted without the Methods' 500 ms rms noise criterion, and (a, ratios) with the adaptation ratios computed here, for figs3a, fig6d and fig3g; neither changes an outcome except (a') for fig2c-subclass-ephys-differences, whose outcome rests on it.",
      "Cell labels (t-type, subclass, layer, region, donor) come from the paper's cell table through the link claim dandiset-cell-link."
     ],
     "assumed": false,
     "disagreed": true,
     "data": "with_inputs",
     "where": null,
     "proposed": null
    },
    {
     "claim": "mouse-cell-link",
     "statement": "The files of DANDI 000020 (version 0.210913.1639) can be matched one to one to the paper's strict mouse L1 cells, and the matched records agree on what both describe.",
     "source": "Methods (mouse data from DANDI 000020); the code's data/mouse_l1_dataset_2023_02_06.csv, mouse_l1_dataset_strict.csv and aibs_mouse_features_E.csv (commit 7777857c1b89b19735e45fd9c9bd7493cc1d805c)",
     "quote": "",
     "statistics": [],
     "kind": "link",
     "inferred": "",
     "outcome": "partially_reproduced",
     "blocked_by": null,
     "notebook": "mouse_comparison.ipynb",
     "notes": "250 of the paper's 272 strict mouse L1 cells have a file in DANDI 000020 version 0.210913.1639; each matches one cell by session id and cell name, and the features computed here agree with the paper's mouse table within 1% for 88 to 100% of cells in most features (f-I slope 58%, adaptation ratios 4 to 21%). 22 cells, 20 of them in VISp layer 1, have no file (91.9% matched); the notebook lists them by specimen id. They include one Lamp5 Krt73 cell (710346584) and 17 LAMP5 cells of VISp layer 1, 12 of them Lamp5 Plch2 Dock5.",
     "deviations": [],
     "assumed": false,
     "disagreed": false,
     "data": "with_inputs",
     "where": null,
     "proposed": null
    },
    {
     "claim": "fig6b-dprime-correlation",
     "statement": "Electrophysiological and transcriptomic d' of human L1 t-type pairs are correlated (Pearson r = 0.59, p = 0.00016).",
     "source": "Results, Cross-modality relationships of L1 subclasses and t-types, paragraph 1; Figure 6B",
     "quote": "d\u2019 metrics were highly correlated between modalities, demonstrating that cell types with distinctive transcriptomes have similarly distinctive electrophysiological properties (Pearson r=0.59, p=0.00016",
     "statistics": [
      {
       "test": "Pearson correlation",
       "statistic": "r = 0.59",
       "p": 0.00016
      }
     ],
     "kind": "finding",
     "inferred": "",
     "outcome": "blocked",
     "blocked_by": "other",
     "notebook": null,
     "notes": "Not attempted in this run; its inputs are public. It needs the patch-seq transcriptomes (ps_human.RData in the paper's code archive) and a refit of the zero-inflated negative binomial discriminant, whose fitted curves (zinb_human.h5) are not public and are refitted as the criterion says; the run gave its time to the claims on the dandiset.",
     "deviations": [],
     "assumed": false,
     "disagreed": false,
     "data": "with_inputs",
     "where": null,
     "proposed": null
    },
    {
     "claim": "fig2d-subclass-morphology",
     "statement": "Human L1 subclasses differ morphologically in vertical orientation of axons and dendrites, axon extent and shape, and dendrite branching (Kruskal-Wallis, FDR p < 1e-3).",
     "source": "Results, Morpho-electric diversity in human L1, paragraph 1; Figure 2D",
     "quote": "Morphologically, subclasses were distinguished by vertical orientation of axons and dendrites, axon extent and shape, and dendrite branching",
     "statistics": [
      {
       "test": "Kruskal-Wallis test",
       "p": 0.001,
       "p_relation": "<",
       "corrected": "FDR"
      }
     ],
     "kind": "finding",
     "inferred": "",
     "outcome": "partially_reproduced",
     "blocked_by": null,
     "notebook": "morphology_tables.ipynb",
     "notes": "64 reconstructed cells of the four core subclasses, 51 features from the paper's deposited morphology table; 4 features have FDR p < 1e-3. Of the three features Fig. 2D plots, the fraction of dendrite below the soma (FDR p = 2.0e-4) and the axon horizontal extent (4.1e-4) are below 1e-3, the number of dendrite branches (0.0034) is not. The axon vertical orientation features have FDR p 0.008 to 0.09.",
     "deviations": [
      "Tested on the paper's per-cell morphology features as deposited in its code archive, not recomputed from the reconstructions in the Brain Image Library.",
      "The claims file gives no criterion; the claim was judged on the three features of Fig. 2D (l1_subclass_distinctness.ipynb cell 21) having FDR p < 1e-3."
     ],
     "assumed": false,
     "disagreed": false,
     "data": "elsewhere",
     "where": "Brain Image Library doi:10.35077/g.606 (reconstructions); morphology features in Zenodo doi:10.5281/zenodo.8310226 (data/RawFeatureWide_human+derivatives.csv)",
     "proposed": null
    },
    {
     "claim": "fig4b-mc4r-axon",
     "statement": "MC4R cells are quantitatively distinct from other L1 types in maximum axonal path distance and axon branch frequency.",
     "source": "Results, Rosehip cells, paragraph 1; Figure 4B",
     "quote": "were quantitatively distinct from other L1 types in terms of maximum axonal path distance and branch frequency",
     "statistics": [
      {
       "test": "Mann-Whitney test against rest of L1",
       "p": 0.0001,
       "p_relation": "<",
       "corrected": "FDR"
      }
     ],
     "kind": "finding",
     "inferred": "",
     "outcome": "partially_reproduced",
     "blocked_by": null,
     "notebook": "morphology_tables.ipynb",
     "notes": "12 MC4R cells against 66. Axon branch frequency 3.01 against 1.36 per 100 um (FDR p = 1.2e-6, below 1e-4). Maximum axonal path 662 against 992 um (FDR p = 0.0029, not below 1e-4).",
     "deviations": [
      "Tested on the paper's deposited morphology features.",
      "Judged on FDR p < 1e-4 for both features, the threshold of the Fig. 4B legend."
     ],
     "assumed": false,
     "disagreed": false,
     "data": "elsewhere",
     "where": "Brain Image Library doi:10.35077/g.606; features in Zenodo doi:10.5281/zenodo.8310226",
     "proposed": null
    },
    {
     "claim": "fig5b-tnfaip8l3-dendrites",
     "statement": "PAX6 TNFAIP8L3 cells have more dendritic branches and a horizontal dendritic extent over 550 um wide.",
     "source": "Results, Bursting PAX6 TNFAIP8L3 cells, paragraph 1; Figure 5B",
     "quote": "as did the number of dendritic branches and large horizontal dendritic extent, over 550 microns wide",
     "statistics": [
      {
       "test": "Mann-Whitney test against rest of L1",
       "p": 0.05,
       "p_relation": "<",
       "corrected": "FDR"
      }
     ],
     "kind": "finding",
     "inferred": "",
     "outcome": "partially_reproduced",
     "blocked_by": null,
     "notebook": "morphology_tables.ipynb",
     "notes": "Two PAX6 TNFAIP8L3 cells against 76. Horizontal dendritic extent 583 and 819 um, both over 550 um, against a median of 140 um (FDR p = 0.017). Dendrite branches 50 and 68 against a median of 32 (FDR p = 0.085, not below 0.05).",
     "deviations": [
      "Tested on the paper's deposited morphology features.",
      "Judged on FDR p < 0.05 (Fig. 5B legend) with TNFAIP8L3 higher for both features, and each cell's extent over 550 um."
     ],
     "assumed": false,
     "disagreed": false,
     "data": "elsewhere",
     "where": "Brain Image Library doi:10.35077/g.606; features in Zenodo doi:10.5281/zenodo.8310226",
     "proposed": null
    },
    {
     "claim": "figs6a-tumor-dendrite-branching",
     "statement": "Cells from tumor patients have higher dendritic branching than cells from epilepsy patients, not explained by brain area or subclass.",
     "source": "Results, Morpho-electric diversity in human L1, paragraph 3; Figure S6A",
     "quote": "with the exception of higher dendritic branching in cells from tumor patients compared to epilepsy patients",
     "statistics": [
      {
       "test": "Mann-Whitney test",
       "p": 0.05,
       "p_relation": "<",
       "corrected": "FDR"
      }
     ],
     "kind": "finding",
     "inferred": "",
     "outcome": "blocked",
     "blocked_by": "metadata_insufficient",
     "notebook": "morphology_tables.ipynb",
     "notes": "The public cell table gives a medical condition for 20 of the 71 reconstructed strict-set cells with an L1 t-type (13 epileptic, 7 not; 51 have none); the code read the conditions of all donors from the Allen Institute's donor database. On these 20 cells the dendrite branch count is higher in non-epileptic donors (median 37 against 26) with FDR p = 0.94, and the check within MTG LAMP5 cells cannot be run, as only epileptic donors are labelled there.",
     "deviations": [],
     "assumed": false,
     "disagreed": false,
     "data": "elsewhere",
     "where": "Brain Image Library doi:10.35077/g.606; morphology features in Zenodo doi:10.5281/zenodo.8310226",
     "proposed": null
    },
    {
     "claim": "fig3b-l1-cytoarchitecture",
     "statement": "Human L1 is thicker than mouse L1 but not less dense, with smaller cell bodies.",
     "source": "Results, Cross-species differences in L1, paragraph 1; Figure 3B and S6D",
     "quote": "human L1 was thicker but not less dense, with smaller cell bodies compared with several mouse neocortical areas",
     "statistics": [
      {
       "test": "Kruskal-Wallis test, Dunn's post hoc (MTG against TEa)",
       "p": 0.05,
       "p_relation": "<",
       "n": "human MTG 76 ROIs (3016 cells, 16 cases); mouse TEa 29 ROIs (400 cells, 3 animals)"
      }
     ],
     "kind": "finding",
     "inferred": "",
     "outcome": "blocked",
     "blocked_by": "metadata_insufficient",
     "notebook": null,
     "notes": "The code (l1_depth_revisions.ipynb cells 8 to 12) rescales the mouse thickness, soma area, density and count values of the deposited source data by each slide's scanner resolution, read from the Allen Institute's LIMS database, which is not public; the deposited CSV does not carry the resolutions. Not run.",
     "deviations": [],
     "assumed": false,
     "disagreed": false,
     "data": "elsewhere",
     "where": "Zenodo doi:10.5281/zenodo.8310226 (data/Source data cortical structure L1.csv and Source data cortical structure L1 all areas.csv)",
     "proposed": null
    },
    {
     "claim": "fig1a-human-types-separated",
     "statement": "In transcriptomic reference data, many human L1 t-types form separated clusters while mouse L1 t-types vary more continuously, suggesting stronger transcriptomic specialization in human L1.",
     "source": "Results, L1 patch-seq pipeline and transcriptomic references, paragraph 1; Figure 1A",
     "quote": "many human t-types formed separated clusters, with others clustered in groups of related t-types, whereas mouse L1 t-types showed more continuous variability",
     "statistics": [],
     "kind": "finding",
     "inferred": "",
     "outcome": "blocked",
     "blocked_by": "other",
     "notebook": null,
     "notes": "Not attempted in this run, not for lack of data; the run gave its time to the claims on the dandiset. It needs the Allen Brain Map human MTG snRNA-seq and mouse VISp scRNA-seq references and the NB discriminant d'.",
     "deviations": [],
     "assumed": false,
     "disagreed": false,
     "data": "elsewhere",
     "where": "Allen Brain Map RNA-seq portal (human MTG snRNA-seq of Hodge et al. 2019, mouse VISp scRNA-seq of Tasic et al. 2018), as the Methods cite them",
     "proposed": null
    },
    {
     "claim": "fig1b-three-subclasses",
     "statement": "Ten human and eight mouse L1 t-types are defined from the references; grouping human t-types with d' below 2.2 forms three subclasses, leaving three t-types ungrouped.",
     "source": "Results, L1 patch-seq pipeline and transcriptomic references, paragraph 1; Figure 1B; Methods, Subclasses and cross-species homology",
     "quote": "we identified 10 L1 t-types in human and 8 in mouse ... This formed three subclasses, with three t-types remaining ungrouped",
     "statistics": [],
     "kind": "finding",
     "inferred": "",
     "outcome": "blocked",
     "blocked_by": "other",
     "notebook": null,
     "notes": "Not attempted in this run, not for lack of data; the run gave its time to the claims on the dandiset. It needs the Allen Brain Map references and the NB discriminant d' grouping.",
     "deviations": [],
     "assumed": false,
     "disagreed": false,
     "data": "elsewhere",
     "where": "Allen Brain Map RNA-seq portal (human MTG snRNA-seq, mouse VISp scRNA-seq)",
     "proposed": null
    },
    {
     "claim": "fig1d-four-homology-subclasses",
     "statement": "Cross-species matching of mouse t-types to human subclasses gives four homology-driven subclasses; two human L1 t-types (SST BAGE2, VIP PCDH20) are homologous to deeper-layer mouse types and are left out.",
     "source": "Results, L1 patch-seq pipeline and transcriptomic references, paragraph 3; Figure 1D",
     "quote": "These matches resulted in four homology-driven subclasses (called subclasses hereafter) with proportions largely comparable across species ... Two additional human L1 t-types (SST BAGE2 and VIP PCDH20) were excluded from cross-species L1 subclasses because of their homology to t-types found in deeper layers, but not L1 of mouse neocortex",
     "statistics": [],
     "kind": "finding",
     "inferred": "",
     "outcome": "blocked",
     "blocked_by": "other",
     "notebook": null,
     "notes": "Not attempted in this run, not for lack of data; the run gave its time to the claims on the dandiset. It needs the references and the integrated coordinates of the cross-species alignment.",
     "deviations": [],
     "assumed": false,
     "disagreed": false,
     "data": "elsewhere",
     "where": "Allen Brain Map RNA-seq portal references and the integrated coordinates in Zenodo doi:10.5281/zenodo.8310226 (data/scalign_coords.rda)",
     "proposed": null
    },
    {
     "claim": "fig1e-subclass-proportions",
     "statement": "Subclass proportions in the L1 patch-seq data differ between human and mouse for every subclass except L1 VIP, PAX6 being the notable difference.",
     "source": "Figure 1E legend; Results, L1 patch-seq pipeline and transcriptomic references, paragraph 3",
     "quote": "All cross-species proportion differences (except L1 VIP) significant at FDR-corrected p<0.001, one vs. rest Fisher\u2019s exact tests",
     "statistics": [
      {
       "test": "one vs. rest Fisher's exact tests",
       "p": 0.001,
       "p_relation": "<",
       "corrected": "FDR"
      }
     ],
     "kind": "finding",
     "inferred": "",
     "outcome": "blocked",
     "blocked_by": "metadata_insufficient",
     "notebook": "mouse_comparison.ipynb",
     "notes": "The test counts, for each species, cells with soma in L1 by category, including the cells of deeper t-types. The human side is in the public cell table (253 cells with soma in L1, 22 of deeper t-types). The mouse side is the code's full mouse patch-seq metadata (shiny_mouse, read from Allen Institute servers); the public mouse table (547 rows) holds only cells of the 8 mouse L1 t-types, none of a deeper t-type, so the mouse denominators and the deeper t-type category cannot be formed. Not run.",
     "deviations": [],
     "assumed": false,
     "disagreed": false,
     "data": "elsewhere",
     "where": "per-cell t-type tables in Zenodo doi:10.5281/zenodo.8310226 (human and mouse l1_dataset_2023_02_06.csv), from the NeMO transcriptomes",
     "proposed": null
    },
    {
     "claim": "figs2a-patchseq-vs-snrnaseq",
     "statement": "Compared with snRNA-seq, patch-seq sampled fewer SST BAGE2 and more PAX6 CDH12 and VIP TSPAN12 cells; MERFISH proportions did not differ from patch-seq.",
     "source": "Results, L1 patch-seq pipeline and transcriptomic references, paragraph 6; Figure S2A",
     "quote": "patch-seq had fewer SST BAGE2 cells and more PAX6 CDH12 and VIP TSPAN12 compared to snRNA-seq (p<0.05, False discovery rate or FDR-corrected Fisher\u2019s exact test). MERFISH had intermediate proportions of SST BAGE2 with no significant differences compared to patch-seq (p>0.18)",
     "statistics": [
      {
       "about": "patch-seq against snRNA-seq",
       "test": "Fisher's exact test",
       "p": 0.05,
       "p_relation": "<",
       "corrected": "FDR"
      },
      {
       "about": "MERFISH against patch-seq",
       "test": "Fisher's exact test",
       "p": 0.18,
       "p_relation": ">"
      }
     ],
     "kind": "finding",
     "inferred": "",
     "outcome": "blocked",
     "blocked_by": "other",
     "notebook": null,
     "notes": "Not attempted in this run, not for lack of data; the run gave its time to the claims on the dandiset. It needs the human MTG snRNA-seq reference counts and the MERFISH tables.",
     "deviations": [],
     "assumed": false,
     "disagreed": false,
     "data": "elsewhere",
     "where": "Zenodo doi:10.5281/zenodo.8310226 (data/merfish_counts.csv, merfish_proportions.csv and the per-cell tables) and the Allen Brain Map human MTG snRNA-seq reference",
     "proposed": null
    },
    {
     "claim": "figs1d-marker-correlation",
     "statement": "Marker gene expression by t-type correlates strongly between patch-seq and the snRNA-seq reference (r = 0.84).",
     "source": "Figure S1D legend; Results, L1 patch-seq pipeline and transcriptomic references, paragraph 5",
     "quote": "Correlation of marker gene expression between patch-seq and snRNA-seq reference (r=0.84, p<10-18)",
     "statistics": [
      {
       "test": "correlation",
       "statistic": "r = 0.84",
       "p": 1e-18,
       "p_relation": "<"
      }
     ],
     "kind": "finding",
     "inferred": "",
     "outcome": "blocked",
     "blocked_by": "other",
     "notebook": null,
     "notes": "Not attempted in this run, not for lack of data; the run gave its time to the claims on the dandiset. It needs the patch-seq transcriptomes and the snRNA-seq reference.",
     "deviations": [],
     "assumed": false,
     "disagreed": false,
     "data": "elsewhere",
     "where": "Zenodo doi:10.5281/zenodo.8310226 (data/ps_human.RData) and the Allen Brain Map human MTG snRNA-seq reference",
     "proposed": null
    },
    {
     "claim": "fig3d-morphology-species",
     "statement": "Human L1 neurites are more tortuous (smaller contraction ratio) than mouse, mouse dendrites are more branched, and mouse axons rarely extend above the dendrites.",
     "source": "Results, Cross-species differences in L1, paragraph 1; Figure 3D left",
     "quote": "Unlike human axons, mouse axons rarely extended above dendrites (Fig. 3D left), perhaps reflecting sublaminar structure found only in the thicker human L1. Human neurites were also structured differently, with smaller contraction ratios (higher tortuosity) compared to the straighter but more heavily branched mouse dendrites",
     "statistics": [
      {
       "test": "2-way ANOVA on ranks, species effect",
       "p": 1e-07,
       "p_relation": "<",
       "corrected": "FDR"
      }
     ],
     "kind": "finding",
     "inferred": "",
     "outcome": "blocked",
     "blocked_by": "other",
     "notebook": null,
     "notes": "Not attempted in this run; its inputs are public (the mouse morphology features are in the code's archive, RawFeatureWide_mouse+derivatives.csv). It needs the mouse morphology table and the code's L1 and L2 border selection; the run gave its time to the claims on the dandiset.",
     "deviations": [],
     "assumed": false,
     "disagreed": false,
     "data": "elsewhere",
     "where": "Brain Image Library doi:10.35077/g.606 and the mouse reconstructions the paper lists (download.brainimagelibrary.org/biccn/zeng/pseq/morph/200526/); features in Zenodo doi:10.5281/zenodo.8310226",
     "proposed": null
    },
    {
     "claim": "figs3c-size-similar",
     "statement": "Human L1 neurons are slightly larger than mouse in horizontal extent, with no species difference in vertical dimensions or dendritic diameter.",
     "source": "Results, Cross-species differences in L1, paragraph 1; Figure S3C",
     "quote": "Although human neurons across subclasses were slightly larger in horizontal extent, no differences were observed in vertical dimensions or dendritic diameter (Fig. S3C)",
     "statistics": [],
     "kind": "finding",
     "inferred": "",
     "outcome": "blocked",
     "blocked_by": "other",
     "notebook": null,
     "notes": "Not attempted in this run; its inputs are public (the mouse morphology features are in the code's archive, RawFeatureWide_mouse+derivatives.csv). It needs the mouse morphology table and the code's L1 and L2 border selection; the run gave its time to the claims on the dandiset.",
     "deviations": [],
     "assumed": false,
     "disagreed": false,
     "data": "elsewhere",
     "where": "Brain Image Library doi:10.35077/g.606; features in Zenodo doi:10.5281/zenodo.8310226",
     "proposed": null
    },
    {
     "claim": "figs3g-morphology-input-resistance",
     "statement": "Passive models built on reconstructed morphologies explain input resistance variability in human (R2 = 0.35) but not mouse (R2 = 0.03) LAMP5 cells.",
     "source": "Results, Cross-species differences in L1, paragraph 3; Figure S3G",
     "quote": "Simulations of passive biophysical models based on reconstructed morphologies showed that input resistance variability in human but not mouse L1 cells can be explained by morphology",
     "statistics": [
      {
       "about": "human",
       "test": "linear regression",
       "statistic": "R2 = 0.35",
       "p": 4e-06
      },
      {
       "about": "mouse",
       "test": "linear regression",
       "statistic": "R2 = 0.03",
       "p": 0.46
      }
     ],
     "kind": "finding",
     "inferred": "",
     "outcome": "blocked",
     "blocked_by": "other",
     "notebook": null,
     "notes": "Not attempted in this run, not for lack of data; the run gave its time to the claims on the dandiset. It needs the reconstructions and passive model simulations.",
     "deviations": [],
     "assumed": false,
     "disagreed": false,
     "data": "elsewhere",
     "where": "Brain Image Library doi:10.35077/g.606 (reconstructions) and the Allen Cell Types model for specimen 475585413 (celltypes.brain-map.org)",
     "proposed": null
    },
    {
     "claim": "figs3e-k-channel-genes",
     "statement": "Genes for A-type K+ channels (KCND2, KCND3, KCNH7) are more highly expressed in mouse than human L1 neurons and correlate with action potential features.",
     "source": "Results, Cross-species differences in L1, paragraph 3; Figure S3E",
     "quote": "the expression of genes (KCND2, KCND3, and KCNH7) associated with fast inactivating, A-type K+ channels ... was higher in mouse neurons and was correlated with several action potential features",
     "statistics": [],
     "kind": "finding",
     "inferred": "",
     "outcome": "blocked",
     "blocked_by": "other",
     "notebook": null,
     "notes": "Not attempted in this run, not for lack of data; the run gave its time to the claims on the dandiset. It needs the human and mouse patch-seq transcriptomes.",
     "deviations": [],
     "assumed": false,
     "disagreed": false,
     "data": "elsewhere",
     "where": "Zenodo doi:10.5281/zenodo.8310226 (data/ps_human.RData, ps_mouse.RData) and NeMO",
     "proposed": null
    },
    {
     "claim": "fig4c-mc4r-genes",
     "statement": "HTR1F marks the rosehip subclass, and GRM5 and RELN are lower in ADARB2 MC4R than in SST CHRNA4 neurons.",
     "source": "Results, Rosehip cells, paragraph 1; Figure 4C",
     "quote": "found markers distinguishing the entire rosehip subclass from the rest of L1, including HTR1F (5HT receptor 1F), along with markers distinguishing the rosehip subtypes ... GRM5 (metabotropic glutamate receptor 5) and RELN (Reelin) showed lower expression in ADARB2 MC4R neurons",
     "statistics": [],
     "kind": "finding",
     "inferred": "",
     "outcome": "blocked",
     "blocked_by": "other",
     "notebook": null,
     "notes": "Not attempted in this run, not for lack of data; the run gave its time to the claims on the dandiset. It needs the human MTG snRNA-seq reference.",
     "deviations": [],
     "assumed": false,
     "disagreed": false,
     "data": "elsewhere",
     "where": "Allen Brain Map human MTG snRNA-seq reference",
     "proposed": null
    },
    {
     "claim": "fig5c-cacna1g",
     "statement": "Both human PAX6 t-types match the alpha7 marker signature and strongly express the T-type calcium channel gene CACNA1G.",
     "source": "Results, Bursting PAX6 TNFAIP8L3 cells, paragraph 3; Figure 5C",
     "quote": "and strongly expressed the T-type calcium channel alpha subunit gene CACNA1G",
     "statistics": [],
     "kind": "finding",
     "inferred": "",
     "outcome": "blocked",
     "blocked_by": "other",
     "notebook": null,
     "notes": "Not attempted in this run, not for lack of data; the run gave its time to the claims on the dandiset. It needs the human MTG snRNA-seq reference.",
     "deviations": [],
     "assumed": false,
     "disagreed": false,
     "data": "elsewhere",
     "where": "Allen Brain Map human MTG snRNA-seq reference",
     "proposed": null
    },
    {
     "claim": "fig6c-subclass-distinctness-species",
     "statement": "Subclasses are more transcriptomically distinct in human than mouse, and in mouse the VIP subclass is highly distinct in all modalities while the others are less so.",
     "source": "Results, Cross-modality relationships of L1 subclasses and t-types, paragraph 1; Figure 6C",
     "quote": "These results confirmed the generally higher transcriptomic distinctness of subclasses in human and show that in mouse the VIP subclass was highly distinct in all modalities, with other subclasses generally less distinct",
     "statistics": [],
     "kind": "finding",
     "inferred": "",
     "outcome": "blocked",
     "blocked_by": "other",
     "notebook": null,
     "notes": "Not attempted in this run, not for lack of data; the run gave its time to the claims on the dandiset. It needs transcriptomes, mouse electrophysiology and morphology d' across modalities.",
     "deviations": [],
     "assumed": false,
     "disagreed": false,
     "data": "elsewhere",
     "where": "Zenodo doi:10.5281/zenodo.8310226 (data/ps_human.RData, ps_mouse.RData), DANDI 000020 and Brain Image Library doi:10.35077/g.606",
     "proposed": null
    }
   ],
   "findings": [
    {
     "kind": "metadata_insufficient",
     "statement": "The files carry no t-type, subclass, cortical layer, brain region or donor medical condition, and no specimen id of the paper's cell table. They can be linked to the paper's table only through the LIMS number in session_id, which equals the Ephys_Roi_Result number inside the table's internal nwb_path column; the dandiset's description points to the code repository but does not say how to link.",
     "found": null,
     "status": null
    },
    {
     "kind": "identifier_missing",
     "statement": "Six files (five of donor 1122692698, one of 1154304730) carry a time stamp such as 2021_08_17_112853 in general/source_script file_name, where the other 204 carry the cell name (for example H21.03.315.11.11.01.01).",
     "found": null,
     "status": null
    },
    {
     "kind": "format_error",
     "statement": "In 145 of 210 files, 2586 long square sweeps that stopped early hold zeros for the samples that were not acquired, where NWB readers such as IPFX expect missing values. IPFX's integrity check passes them and reads the step to 0 mV as a spike, which changes rheobase and spike features unless the sweeps are left out.",
     "found": null,
     "status": null
    },
    {
     "kind": "format_error",
     "statement": "bias_current is stored in pA, with a unit attribute of pA, where the NWB schema fixes the unit of bias_current as amperes; a reader that follows the schema gets values 10^12 times too large.",
     "found": null,
     "status": null
    },
    {
     "kind": "metadata_error",
     "statement": "The two files of donor 922967568 state an age of P0.0D where the paper's table gives the age as unknown, and one file (session 977186103, cell H19.03.321.12.04.01.04) states sex unknown where the table gives M. Ages are stated as whole years times 365 days. The species is written as Homo Sapiens in 192 files and Homo sapiens in 18.",
     "found": null,
     "status": null
    },
    {
     "kind": "data_missing",
     "statement": "The dandiset holds only the 210 Seattle cells of the paper's primary set. The 50 Amsterdam (22) and Szeged (28) cells of the strict L1 set, which the paper uses as held-out validation (Fig. S8B) and in its LAMP5 t-type analysis, are not in it, and the paper names no other deposit of them.",
     "found": null,
     "status": null
    },
    {
     "kind": "selection_undocumented",
     "statement": "The dandiset holds exactly the paper's primary set P, the Seattle cells of the strict L1 set (soma in L1 or within 40 um of the L1/L2 border) whose long square features were extracted. The 12 Seattle cells of the strict set that failed feature extraction and the other Seattle cells of the paper's table are not deposited, and the dandiset's description does not say how the cells were selected.",
     "found": null,
     "status": null
    },
    {
     "kind": "metadata_insufficient",
     "statement": "The donor medical condition is given in the public cell table for only 86 of the 194 cells of P in L1 t-types and 20 of the 71 reconstructed strict-set cells with an L1 t-type; the paper's code read it from the Allen Institute's donor database. The claims on medical condition cannot be repeated on the paper's cells.",
     "found": null,
     "status": null
    },
    {
     "kind": "methods_unavailable",
     "statement": "Several features the claims rest on (irregularity ratio, depolarizing sag, ADP slope and voltage, sag time to peak and time constant, the ramp, short square and chirp features) come from code the Methods say will be released in a future IPFX version; it is not in IPFX 2.1.2 nor in the code repository. The adaptation ratios computed here from the Methods' definition (third over first spike on the lowest step with at least 4 spikes) agree with the paper's values within 1% for only 4 to 53% of cells (six ratios; 4 to 37% for the five not used). The authors are asked whether the feature code can be released.",
     "found": null,
     "status": null
    },
    {
     "kind": "methods_unavailable",
     "statement": "The authors are asked what d' counts as high in Fig. 6A, where LAMP5 NMBR against LCP2 is said to be the only within-subclass pair with high d'. With the 2.2 grouping threshold it holds here on the features computed here for three seeds and on the paper's values for one of three; with the median between-subclass d' (2.40 to 2.62) no within-subclass pair is high.",
     "found": null,
     "status": null
    },
    {
     "kind": "methods_unavailable",
     "statement": "The authors are asked whether the two-way ANOVA of Fig. 3D was run on ranks with Type II sums of squares and heteroskedasticity-corrected tests (legend and Data S4), or on raw values with Type I sums of squares, as the code calls anova_lm without its robust option. With every visual area, as the code takes the mouse cells, the AP height species effect is FDR p = 2.6e-38 under the first and 1.4e-43 under the second, on either side of the legend's 1e-38.",
     "found": null,
     "status": null
    },
    {
     "kind": "methods_unavailable",
     "statement": "The authors are asked how the sweep QC criterion of rms noise below 0.5 mV over 500 ms (Methods, Quality control) was measured. Measured with IPFX on the deposited files, it leaves out sweeps from which the paper's per-cell values were computed: in 7 of the 9 cells where IPFX falls back to a lower sweep at rheobase + 40 pA and the value disagrees with the paper's, the paper's value comes from a sweep that fails only this criterion here. Without it, 38 of the 39 recomputed features agree with the paper's values for more cells, and the initial firing rate difference across subclasses (Fig. 2C) has FDR p 3.9e-9 against 4.7e-7 with it.",
     "found": null,
     "status": null
    },
    {
     "kind": "code_mismatch",
     "statement": "The Fig. S3A legend gives R2 = 0.74 for the upstroke adaptation ratio against sag. The code's regression (run_regressions, OLS with HC3) on the paper's own per-cell values over the cells of P in L1 t-types gives R2 = 0.36 (192 cells), and 0.31 on t-type medians.",
     "found": null,
     "status": null
    },
    {
     "kind": "other",
     "statement": "For the mouse comparison (DANDI 000020 version 0.210913.1639, an outside input), 22 of the paper's 272 strict mouse L1 cells, 20 of them in VISp layer 1 and one of them a Lamp5 Krt73 cell, have no file in that version, and none of the paper's 15 mouse TEa cells does.",
     "found": null,
     "status": null
    },
    {
     "kind": "code_mismatch",
     "statement": "The paper compares human L1 with mouse VISp L1 (Fig. 3E legend; Methods), while the code's cross-species notebook (l1_cross_species.ipynb cell 5) takes mouse cells from every visual area (VISp, VISpm, VISl and others; 185 cells against 150 in VISp here). With VISp alone the AP height species effect is FDR p 3.3e-34 to 3.4e-33, above the legend's 1e-38.",
     "found": null,
     "status": null
    },
    {
     "kind": "other",
     "statement": "Nucleated-patch voltage clamp sweeps are in 38 human files of the dandiset and 25 strict mouse L1 files of DANDI 000020, but the patch surface areas the paper's conductance densities (Fig. 3F) need, measured from images, are not deposited.",
     "found": null,
     "status": null
    }
   ],
   "reconciliation": null,
   "verifications": [
    {
     "date": "2026-10-31",
     "kind": "agent",
     "by": "Independent agent run by CatalystNeuro (same harness, separate context, did not write the notebooks)",
     "reviewed": "Round 1. The three executed notebooks with the conclusion cells and the summaries removed, the helper module and the claims file, with the paper. The verifier ran the helper statistics on synthetic data and queried the archive; it gave its own outcome for each claim. It disagreed with five outcomes (dandiset-cell-link, fig4b-adarb2-irregular-firing, fig6a-ephys-dprime-subclass-structure, fig4d-mouse-no-rosehip-ephys and fig5d-mouse-pax6-doublets) and corrected several reasons for blocks. Report in transcripts/000630/verifier-1-report.md.",
     "agreed": 36,
     "of": 41,
     "disagreements": [
      {
       "claim": "dandiset-cell-link",
       "outcome": "partially_reproduced",
       "reason": "R c14: all 210 files match 210 distinct rows. Sex agrees for 209/210 (one file says unknown). Age agrees for 207/208, with one off by 1 year under the notebook's own floor(x+0.05) rounding. 39 of 44 computed features agree within 1% for at least half the cells. The criterion asks every record to agree, so this is partial (>=90%). Also (B11) the age match uses an ad hoc rounding, and (B2) the cells where first_isi_inv_hero disagrees are never inspected."
      },
      {
       "claim": "fig4b-adarb2-irregular-firing",
       "outcome": "not_reproduced",
       "reason": "R c38: (i) fails, because ADARB2 against CHRNA4 on ir_late_mean has Dunn FDR 0.18. (ii) fails, because CHRNA4 is above LAMP5 with FDR 0.046. (iii) and (iv) hold under both readings of faster."
      },
      {
       "claim": "fig6a-ephys-dprime-subclass-structure",
       "outcome": "partially_reproduced",
       "reason": "R c42: mean within-subclass d' is below between-subclass d' in all 6 runs. NMBR/LCP2 is the only high within-subclass pair under >=2.2 in 4 of 6 runs (only seed 0 under (b)). Under above median between no within pair is high, so the criterion's both readings condition fails."
      },
      {
       "claim": "fig4d-mouse-no-rosehip-ephys",
       "outcome": "not_reproduced",
       "reason": "M c12: MC4R sag is above LAMP5's. But only 15 of 30 irregular cells are LAMP5 (not most), and the largest group is Lamp5 Plch2 Dock5 (8), not Ntn1 Npy2r (6). Both fail under both readings. 11 L1 VIP cells are irregular. The mouse tables are built only from the 250 cells with files, even under reading (b) (B5)."
      },
      {
       "claim": "fig5d-mouse-pax6-doublets",
       "outcome": "not_reproduced",
       "reason": "M c14: PAX6's sag_depol median is the highest, but its first_isi_inv_hero median (97.7, n=2) is below L1 VIP's (171). The missing fourth Lamp5 Krt73 cell is unexplained (B5)."
      }
     ]
    },
    {
     "date": "2026-10-31",
     "kind": "agent",
     "by": "Independent agent run by CatalystNeuro (same harness, separate context, did not write the notebooks)",
     "reviewed": "Round 2. The full executed notebooks, with the conclusion cells, the summaries and the attempt record, the helper module and the claims file, with the paper. The verifier agreed with 39 outcomes and disagreed with two (fig2c-lamp5-gradient and fig6a-ephys-dprime-subclass-structure, which it would give as not_reproduced), asked for the sweep at rheobase + 40 pA to be checked for fig2c-subclass-ephys-differences, reported a possible misreading for fig5d-mouse-pax6-doublets, and listed sentences the outputs do not support. Report in transcripts/000630/verifier-2-report.md.",
     "agreed": 39,
     "of": 41,
     "disagreements": [
      {
       "claim": "fig2c-lamp5-gradient",
       "outcome": "not_reproduced",
       "reason": "Neither criterion names a partial level. The record changed fig4b-adarb2, fig4d and fig5d to not_reproduced on exactly that ground, so this claim should follow the same rule; sag misses 0.001 under every reading."
      },
      {
       "claim": "fig6a-ephys-dprime-subclass-structure",
       "outcome": "not_reproduced",
       "reason": "Neither criterion names a partial level, and the record changed fig4b-adarb2, fig4d and fig5d to not_reproduced on exactly that ground, so this claim should follow the same rule; the second reading of high fails."
      }
     ]
    },
    {
     "date": "2026-10-31",
     "kind": "agent",
     "by": "Independent agent run by CatalystNeuro (same harness, separate context, did not write the notebooks)",
     "reviewed": "Round 3. Only the three proposed changes of outcome and the analysis behind them in the executed notebook, with the claims file and the paper. The verifier agreed with all three, for fig2c-subclass-ephys-differences on the condition that the assumption be recorded as a structured field. Report in transcripts/000630/verifier-3-report.md.",
     "agreed": 3,
     "of": 3,
     "disagreements": []
    }
   ]
  }
 ],
 "entries": [
  {
   "name": "000630-chartrand-2023",
   "type": "reproduction",
   "title": "Human layer 1 interneuron electrophysiology (Chartrand et al. 2023)",
   "description": "Tests the electrophysiology findings of Chartrand et al. (2023) on dandiset 000630, which holds the current clamp recordings of the paper's 210 Seattle human L1 patch-seq cells. Features are extracted with IPFX from the long square sweeps and compared with the paper's per-cell table; cell types come from the paper's cell table, linked by the LIMS identifier. A second notebook compares the human cells with the paper's mouse L1 cells from dandiset 000020. A third notebook tests four morphology claims on the paper's deposited morphology features.",
   "datasets": [
    "dandi:000020",
    "dandi:000630"
   ],
   "papers": [
    "10.1126/science.adf0805"
   ],
   "derived_from": [],
   "notebooks": [
    {
     "path": "reproduce_chartrand_2023.ipynb",
     "title": "Reproducing the human electrophysiology findings of Chartrand et al. (2023) from dandiset 000630",
     "executed": "notebooks/000630-chartrand-2023/reproduce_chartrand_2023.html"
    },
    {
     "path": "mouse_comparison.ipynb",
     "title": "Human against mouse L1 cells, with dandiset 000020",
     "executed": "notebooks/000630-chartrand-2023/mouse_comparison.html"
    },
    {
     "path": "morphology_tables.ipynb",
     "title": "Morphology claims from the paper's deposited feature table",
     "executed": "notebooks/000630-chartrand-2023/morphology_tables.html"
    }
   ]
  }
 ]
}