Claims 45
From the paper Human interneuron patch-seq electrophysiology · claims file · checked 2026-10-23: ready
- cell-metadata-link
The dandiset's NWB files can be matched to the paper's per-cell metadata and feature tables in its code repository, and features recomputed from the files agree with those tables.
How it will be judged
Files are matched to rows of LeeDalley_manuscript_metadata_v2.csv by its nwb_id column (or by specimen id where the NWB metadata carry it). The link holds (reproduced) when at least 90% of the rows with Has_ephys true are found in the dandiset, and, for the matched cells, rheobase_i, input_resistance, sag and width_rheo computed here with IPFX are within 10% of the LeeDalley_ephys_fx.csv values for at least 90% of the cells with both values. It is partially reproduced when the matching holds and the features agree for 50 to 90% of cells, and not reproduced otherwise. The number of NWB files without a metadata row is reported.
- dataset-composition
The study characterized 778 human GABAergic neurons in layers 2 to 6, covering 44 of the 45 GABAergic t-types, with electrophysiology from 75 LAMP5/PAX6, 99 VIP, 149 SST and 352 PVALB neurons.
properties of 778 neurons from the four canonical GABAergic subclasses in human cortex. We described neurons in cortical layers 2 to 6 (L2–6) across 44 transcriptomic types ... LAMP5/PAX6 ( n = 75), VIP ( n = 99), SST ( n = 149), and PVALB ( n = 352)
Results, Comparison of features across human GABAergic subclasses, paragraphs 1 and 3How it will be judged
Counted over the dandiset's NWB files that pass the cell inclusion criteria of the Methods and have at least one QC-passing long square sweep, labelled through cell-metadata-link. The subclass counts are read two ways, as the text does not say which is meant, (a) Revised_subclass_label (SST FRZB in PVALB, as for subclass-level analyses) or (b) subclass_label. The 778 and 44 are counted over all labelled cells, as they include cells without electrophysiology; that the dandiset holds fewer cells than 778 is expected and reported. Reproduced when the four subclass counts are within tolerance under one reading.
- fig1b-more-genes-in-culture
Patch-seq neurons from slices in culture (2 to 16 days) had more genes detected than neurons from acute slices.
Patch-seq neurons from culture preparations recorded at 2 to 16 days in culture (DIC) had significantly ( P < 0.0001) higher genes detected as compared with the acute paradigm
Results, Morphological, electrophysiological, and transcriptomic features in the slice culture paradigm, paragraph 1; Figure 1BHow it will be judged
Genes detected per cell, culture against acute (condition column), two-sided Mann-Whitney test. Reproduced when the culture median is higher at p < 0.0001.
- figs1-microglial-signature
Culture neurons fall into two clusters, with and without a microglial gene signature, and the signature is also present, less strongly, in a subset of acute neurons.
These neurons clustered into two domains, with or without microglial gene expression signatures ... This microglial signature ( data S2 ) was also observed in a subset of neurons collected with the acute paradigm, although the effect was less pronounced
Results, Morphological, electrophysiological, and transcriptomic features in the slice culture paradigm, paragraph 1; fig. S1B and S1CHow it will be judged
As in Methods (Assigning cells a microglial signature) and the notebook (chunk "identify immune signature cells in culture and acute"), Seurat clustering of culture cells at resolution 0.001 into two clusters, then of culture and acute cells separately on the 269 culture microglial genes at resolution 0.1. Reproduced when the culture cells form two clusters, one with high microglial gene expression, and some acute cells are called microglial, with the effect less pronounced in acute cells. "Less pronounced" is not defined; its readings are (a) a lower fraction of acute than of culture cells called microglial and (b) lower mean expression of the 269 genes in acute than in culture microglial cells. The outcome is the one both give; blocked as methods_underspecified when they differ.
- fig1d-nms-no-difference
The normalized marker sum score does not differ between acute and culture neurons.
we found no difference between the acute and culture paradigms ( Fig. 1D )
Results, Morphological, electrophysiological, and transcriptomic features in the slice culture paradigm, paragraph 1; Figure 1DHow it will be judged
A claim of absence, judged on the paper's test, a two-sided Mann-Whitney test of NMS, acute against culture (Fig. 1D legend). Reproduced when p >= 0.05; the medians and the effect size are reported either way.
- fig1g-wfdc2-morphology-unchanged
Fifty morphological features of depth-matched PVALB WFDC2 neurons do not differ between culture (n = 9) and acute (n = 10) slices.
We also evaluated 50 morphological features of cortical depth–matched PVALB WFDC2 neurons ( Fig. 1F ) from culture ( n = 9) and acute paradigms ( n = 10) and didn’t find any difference among the two conditions
Results, Morphological, electrophysiological, and transcriptomic features in the slice culture paradigm, paragraph 2; Figure 1G; fig. S2AHow it will be judged
A claim of absence, judged on the paper's test, a Mann-Whitney U test per feature with Benjamini-Hochberg FDR correction across the 50 features (Methods, Statistical analysis of variability; preprint, P > 0.05). Reproduced when no feature has FDR-corrected p < 0.05. Which 9 and 10 cells were depth-matched is not given; readings are (a) the cells with a reconstruction in the culture and acute groups, if that gives 9 and 10, and (b) those cells after matching soma depth, the method of matching being open.
- figs2b-wfdc2-ephys-43-of-84
In PVALB WFDC2 neurons, 43 of 84 electrophysiological features differ between the acute and culture paradigms.
Analysis of electrophysiological responses from PVALB WFDC2 neurons revealed that 43 of 84 features differed across the two paradigms
Results, Morphological, electrophysiological, and transcriptomic features in the slice culture paradigm, paragraph 2; fig. S2B; preprint, same sectionHow it will be judged
PVALB WFDC2 cells (Transcriptomic_type) with electrophysiology, acute against culture (condition). Per feature a two-sided Mann-Whitney test, Benjamini-Hochberg FDR across features, counted at FDR-corrected p < 0.05 (preprint, "P < 0.05, false discovery rate (FDR)-corrected Mann–Whitney test"). The feature set is reading (a), the 84 non-chirp columns of LeeDalley_ephys_fx.csv; the count over all 93 columns is shown as a variant and does not decide the outcome, as the paper counts 84 features. Whether the microglial groups are pooled is not said; they are pooled, as condition has only acute and culture, and the split by condition2 is shown as a variant.
- fig1l-width-sag-differ
In PVALB WFDC2 neurons AP width and sag differ between acute and culture slices (sag is higher in culture), while interspike interval and input resistance do not.
AP width and sag are two of the key features that were different between the two groups ( Fig. 1L ); however, multiple other cardinal intrinsic membrane properties, such as interspike interval or input resistance, were not
Results, Morphological, electrophysiological, and transcriptomic features in the slice culture paradigm, paragraph 2; Figure 1L; preprint Discussion, paragraph 2 (direction of sag)How it will be judged
PVALB WFDC2 cells, acute against culture, two-sided Mann-Whitney tests with Benjamini-Hochberg FDR across the features of figs2b-wfdc2-ephys-43-of-84 (reading a). Reproduced when sag and AP width each differ at FDR-corrected p < 0.0001 with sag higher in culture, and ISI and input resistance (the absence part, judged on the same test) have FDR-corrected p >= 0.05. AP width is read as width_rheo, width_hero, width_short_square or width_ramp, and ISI as mean_isi_rheo or mean_isi_hero; the outcome is the one all readings give, and the claim is blocked as methods_underspecified when they differ. The direction of the width change is reported.
- figs2c-no-change-with-days
The features that differ in culture already differ at 2 days in culture and do not change with further days in culture.
The features that were different between the two paradigms were observed at the earliest time point (2 DIC) and did not change with additional days in culture
Results, Morphological, electrophysiological, and transcriptomic features in the slice culture paradigm, paragraph 2; fig. S2C (preprint Supplementary Figure 2C, sag and AP width against days in culture)How it will be judged
A claim of absence, judged on the test the Methods give for continuous variables, Pearson's correlation of sag and of AP width with Days_In_culture over culture PVALB WFDC2 cells (the two features the preprint's Supplementary Figure 2C plots), FDR-corrected across the two. Reproduced when neither correlation is significant at FDR-corrected p < 0.05 and the 2 DIC cells already differ from the acute median in the direction of the culture group. The same test over all features found in figs2b-wfdc2-ephys-43-of-84 is shown as a variant. The AP width readings are those of fig1l-width-sag-differ.
- fig1i-wfdc2-ephys-umap-overlap
In electrophysiology UMAP space of all 84 features, acute and culture PVALB WFDC2 neurons occupy a similar space.
When projecting all 84 electrophysiological features through the UMAP, PVALB WFDC2 neurons from both paradigms occupied a similar space
Results, Morphological, electrophysiological, and transcriptomic features in the slice culture paradigm, paragraph 2; Figure 1IHow it will be judged
The paper gives no UMAP parameters and no overlap measure, so one is fixed here. All cells with electrophysiology, features of reading (a) z-scored with missing values set to the feature median; for each acute PVALB WFDC2 cell, the fraction of its 15 nearest neighbours (Euclidean, in feature space and in a umap-learn embedding with default parameters) among the PVALB WFDC2 cells that are culture cells, and the same for culture cells. Reproduced when both groups' mean cross-group fraction is at least half of what random labels with the same group sizes give (median of 1000 shuffles), in both spaces. The neighbour count and the one-half threshold are choices, not the paper's; the fractions are reported against the shuffle distribution and for 10 and 30 neighbours.
- figs2d-seven-channel-genes
Of 114 voltage-gated ion-channel genes, only seven differ in expression between culture and acute neurons.
We examined 114 voltage-gated ion-channel genes and found only seven that were different in culture versus acute paradigm
Results, Morphological, electrophysiological, and transcriptomic features in the slice culture paradigm, paragraph 3; fig. S2D; Discussion, paragraph 2How it will be judged
The Methods give no test; the notebook settles it (chunks "linear model gene expression WFDC2" and "fold change and ion channels WFDC2"). PVALB WFDC2 cells, genes kept by the notebook with mean expression > 0.01, a linear model expression ~ condition + isImmune + condition x isImmune per gene, p-values of the condition term adjusted with R's p.adjust default (Holm, though the code comment says FDR), channel genes from vgic_genes.csv, counted at adjusted p <= 0.01. Reproduced when the count is 7 and the channel genes tested number about 114 (within 10%). The notebook's all-subclass model is a variant.
- fig2b-ephys-umap-subclasses
The four GABAergic subclasses are separable in electrophysiology feature space, with PVALB neurons occupying the most distinct region.
PVALB, SST, VIP, and LAMP5/PAX6 interneurons were separable when visualized by electrophysiological ( Fig. 2B ) and morphological ( Fig. 2C ) features in UMAP space, with PVALB neurons occupying the most distinct subregions in both UMAPs
Results, Comparison of features across human GABAergic subclasses, paragraph 1; Figure 2BHow it will be judged
No UMAP parameters or separability measure are given, so one is fixed here. Features of reading (a), z-scored, missing values set to the median; for each cell the fraction of its 15 nearest neighbours (in feature space and in a default umap-learn embedding) of its own subclass. Reproduced when every subclass's mean same-subclass fraction exceeds its share of all cells by at least 0.2, and PVALB has the highest mean same-subclass fraction, in both spaces.
- fig2c-morph-umap-subclasses
The four subclasses are separable in morphological feature space, PVALB the most distinct.
PVALB, SST, VIP, and LAMP5/PAX6 interneurons were separable when visualized by electrophysiological ( Fig. 2B ) and morphological ( Fig. 2C ) features in UMAP space
Results, Comparison of features across human GABAergic subclasses, paragraph 1; Figure 2CHow it will be judged
As for fig2b-ephys-umap-subclasses, on the morphology features of the 140 reconstructed neurons.
- fig2e-lamp5-shortest-extent
LAMP5/PAX6 neurons have the shortest soma-to-branch-tip Euclidean distance for both axon and dendrite.
They also had the shortest soma–to–branch tip Euclidean distance for both axon and dendrite
Results, Comparison of features across human GABAergic subclasses, paragraph 2; Figure 2EHow it will be judged
axon_max_euclidean_distance and basal_dendrite_max_euclidean_distance by subclass, Kruskal-Wallis with Dunn's post hoc, FDR-corrected (Fig. 2E legend). Reproduced when the LAMP5/PAX6 median is the lowest of the four for both and Kruskal-Wallis is significant at p < 0.05.
- fig2f-ap-kinetics
APs of the PVALB subclass have the fastest kinetics, and LAMP5/PAX6 APs have the highest peak amplitude of the four subclasses.
APs from PVALB subclass had the fastest kinetics, whereas LAMP5/PAX6 had the highest peak amplitude versus other subclasses
Results, Comparison of features across human GABAergic subclasses, paragraph 3; Figure 2FHow it will be judged
Figure 2F shows APs to a current pulse, a 3 ms pulse in the preprint, so short square features are the first reading; rheobase-sweep features are the second. Kinetics is read as AP width (width_short_square or width_rheo) and as upstroke/downstroke ratio; peak as peak_deltav (peak relative to threshold) of the same sweep. Kruskal-Wallis across the four subclasses with Dunn's post hoc, FDR-corrected (the tests of Fig. 2H). The ratio is peak upstroke over peak downstroke (Methods), so faster repolarization gives a lower ratio. Reproduced when PVALB has the narrowest width and the lowest upstroke/downstroke ratio, and LAMP5/PAX6 the highest peak, each with Dunn's p < 0.05 against every other subclass, under all readings; blocked as methods_underspecified when readings disagree.
- fig2g-sag-by-subclass
LAMP5/PAX6 and VIP neurons have minimal sag compared with PVALB and SST neurons.
LAMP5/PAX6 and VIP subclasses had minimal sag compared with PVALB and SST subclasses
Results, Comparison of features across human GABAergic subclasses, paragraph 3; Figure 2GHow it will be judged
Sag as IPFX computes it at the hyperpolarizing sweep closest to -100 mV (Methods), by subclass, Kruskal-Wallis with Dunn's post hoc, FDR-corrected. Reproduced when the LAMP5/PAX6 and VIP medians are both below the PVALB and SST medians and all four of those pairs differ at Dunn's p < 0.05.
- fig2h-several-features-distinct
Several additional electrophysiological features differ between the subclasses.
Several additional features further demonstrated the electrophysiological distinctness of each subclass
Results, Comparison of features across human GABAergic subclasses, paragraph 3; Figure 2HHow it will be judged
The features plotted in Figure 2H are not named in any text read. Kruskal-Wallis across subclasses for every feature of reading (a), Benjamini-Hochberg FDR across features. "Several" is read as at least three. Reproduced when at least three features other than width, upstroke/downstroke ratio, peak and sag have FDR-corrected p < 0.05 and at least one Dunn's pair significant at p < 0.05. The count of significant features is reported.
- fig3c-frzb-ephys-with-pvalb
Most SST FRZB neurons colocalize with the PVALB subclass in electrophysiology space.
Examination of all 84 electrophysiological features, visualized using a UMAP, revealed that most of the SST FRZB neurons colocalized with the PVALB subclass
Results, Revising subclass assignment based on multimodal data, paragraph 2; Figure 3CHow it will be judged
Groups are SST (subclass_label SST without SST FRZB), PVALB (without SST FRZB) and SST FRZB (Transcriptomic_type). No UMAP parameters are given, so a measure is fixed here. Features of reading (a), z-scored over SST and PVALB cells; for each SST FRZB cell, its 15 nearest SST or PVALB neighbours (feature space, and a default umap-learn embedding). Reproduced when more than half of the SST FRZB cells have a majority of PVALB neighbours in both spaces.
- fig3e-frzb-ap-like-pvalb
The AP upstroke/downstroke ratio and AP width of SST FRZB neurons align with fast-spiking PVALB neurons and differ from SST neurons.
We observed the AP kinetics, such as upstroke/downstroke ratio, AP width, and AP firing pattern of SST FRZB neurons aligned with fast-spiking interneurons of the PVALB subclass
Results, Revising subclass assignment based on multimodal data, paragraph 2; Figure 3D, 3E and 3IHow it will be judged
Groups as in fig3c-frzb-ephys-with-pvalb. Kruskal-Wallis over the three groups with Dunn's post hoc, FDR-corrected (Fig. 3E legend). Reproduced when, for both features, SST FRZB differs from SST at Dunn's p < 0.0001 and the SST FRZB median is nearer the PVALB median than the SST median. Which pairs carry the legend's asterisks is not said; the SST FRZB against PVALB result is reported. Features are read as upstroke_downstroke_ratio and width of the rheobase sweep, the short square and the rheobase + 40 pA sweep; the outcome is the one all readings give.
- figs6-sst-features-acute-culture
The upstroke/downstroke ratio and AP width of SST subclass neurons do not differ between acute and culture recordings.
Additionally, these select features of the SST subclass were not different in recordings from acute versus culture paradigms
Results, Revising subclass assignment based on multimodal data, paragraph 2; fig. S6How it will be judged
A claim of absence, judged on the test used for acute against culture elsewhere in the paper (FDR-corrected Mann-Whitney, Fig. 1 legend), since the fig. S6 legend is missing. SST subclass cells (subclass_label SST, read with and without SST FRZB), the two features of fig3e-frzb-ap-like-pvalb under its sweep readings. Reproduced when no feature differs at FDR-corrected p < 0.05 under every reading; the effect sizes are reported.
- fig3-frzb-transcriptome-intermediate
SST FRZB neurons express SST higher than PVALB but lack several key genes of each subclass while expressing several others, and most SST FRZB Patch-seq samples lie with the PVALB subclass in transcriptomic UMAP space.
Whereas neurons belonging to the SST FRZB transcriptomic type did express SST higher than PVALB, they lacked expression of multiple key genes that define each subclass ... the vast majority of SST FRZB samples localized to the PVALB subclass within the transcriptomic UMAP space in our Patch-seq dataset
Results, Revising subclass assignment based on multimodal data, paragraphs 1 and 2; Figure 3A and 3BHow it will be judged
Top 20 genes differentially expressed between all PVALB and all SST types (excluding SST FRZB) in the snRNA-seq data (Methods); SST against PVALB expression in the SST FRZB nuclei of the snRNA-seq data (Fig. 3A legend, a heat map of snRNA-seq data); for the UMAP, the integrated space of the Methods (Visualization of integrated transcriptomic space; the notebook's chunk "run Seurat integration by condition"). Reproduced when the SST FRZB median SST expression exceeds its PVALB expression, at least two subclass genes of each list are not expressed (median CPM 0) and at least two are (median CPM > 0), thresholds the paper does not give, and more than half of the SST FRZB Patch-seq cells have a majority of PVALB among their 15 nearest SST or PVALB neighbours.
- fig3g-frzb-morphology-like-pvalb
SST FRZB neurons overlap more with PVALB than with SST neurons in morphology space, their horizontal axon extent and axon versus dendrite earth mover's distance resemble PVALB basket cells, and in co-clustering they fall more often into PVALB-dominated types.
Morphological features such as horizontal axon extent and a measure of the dissimilarity of the axon and dendrite compartments (axon versus dendrite earth mover’s distance) showed that SST FRZB neurons had individual morphological features that were more similar to PVALB basket cells than to SST neurons
Results, Revising subclass assignment based on multimodal data, paragraph 3; Figure 3F to 3HHow it will be judged
axon_extent_x and axon_emd_with_basal_dendrite for SST, SST FRZB and PVALB, Kruskal-Wallis with Dunn's post hoc, FDR-corrected (Fig. 3G legend); co-clustering as the Methods give it (95% subsamples, shared nearest neighbour graph, fast-greedy community detection, 500 resamplings, Ward linkage on the co-clustering matrix). The features entering the graph, the number of neighbours of the graph and where the Ward tree is cut are not given; the paper names types 1 to 6, so the tree is cut at six clusters, and the features are all columns of LeeDalley_morpho_features.csv; these are choices, and 10 and 20 graph neighbours are both run. A cluster is PVALB- or SST-dominated by its majority subclass. Reproduced when the SST FRZB medians are nearer the PVALB than the SST medians for both features, SST differs from SST FRZB at p < 0.01 for at least one (the legend does not say which pairs carry the asterisks; the preprint gave P > 0.05 for this panel, the version of record gives ** P < 0.01, and the version of record is followed), and more SST FRZB cells fall in PVALB-dominated clusters than in SST-dominated ones under both neighbour counts.
- fig4a-calb1-morphology-types
The SST CALB1 t-type contains four qualitative morphology types, Martinotti cells, double bouquet cells, one basket-like cell and sparse SST cells.
In humans, we qualitatively identified four morphological types within SST CALB1. Three of these types resembled previously described neurons
Results, Morphological heterogeneity within SST CALB1, paragraph 1; Figure 4A and 4BHow it will be judged
The types were assigned by eye; the test is that the reconstructions labelled with each type in the metadata show the defining traits the Results give. Reproduced when MCs have more axon length in L1 than below their soma layer, DBCs have the narrowest axon_extent_x relative to dendrite width (axon_over_basal_dendrite_extent_x) of the four, the basket-like cell has the least L1 axon, and sparse SST cells have fewer axon branches per length than MCs.
- fig4c-calb1-ephys-by-morphology
Within SST CALB1, sparse SST neurons tend to have lower sag, higher adaptation ratio and slightly delayed AP onset than DBCs and MCs, and MCs have the highest AP peak.
sparse SST neurons show a lower degree of sag, a higher adaptation ratio, and a slightly delayed onset of AP firing as compared with DBCs and MCs, whereas the peak of the AP is highest in MCs
Results, Morphological heterogeneity within SST CALB1, paragraph 2; Figure 4CHow it will be judged
SST CALB1 cells with Qualitative_morphology_type MC, DBC, sparse SST or basket-like, with electrophysiology. The paper calls these trends from few neurons and does not say which pairs are significant, so the claim is judged on medians, with Kruskal-Wallis and Dunn's FDR-corrected results reported. Sag is sag; adaptation ratio is the Methods' ISI adaptation ratio (first to third spike, isi_adapt_ratio), as the Methods keep "adaptation ratio" for these first-to-third spike ratios and call adapt_rheo and adapt_hero the adaptation index, which is shown as a variant only; onset as latency_rheo and latency_hero; peak as peak_deltav_rheo and peak_deltav_hero. Reproduced when, under every reading, the sparse SST median sag is below, its adaptation and latency medians above, those of DBC and MC, and the MC peak median is the highest of the four types; partially reproduced when three of the four statements hold.
- fig4d-calb1-genes-by-morphology
A UMAP on 253 genes differentially expressed between DBC, MC and sparse SST neurons splits SST CALB1 neurons into three clusters that largely separate the morphology types.
With this variable gene set, three distinct clusters appeared with morphological types largely separated across clusters
Results, Morphological heterogeneity within SST CALB1, paragraph 3; Figure 4DHow it will be judged
The Methods (Calculating differentially expressed genes, case iv) give Seurat FindMarkers, t test, min.pct 0, logfc.threshold 0, Bonferroni, between any pair of DBC, MC and sparse SST cells; the genes are the union over the three pairs. The notebook (chunks "morpho of immune culture cells" and "morpho of immune culture cells dex", read at 5eb357a) instead runs FindAllMarkers (t test, logfc.threshold 0) on SST CALB1 culture_immune cells of known morphology, after dropping immune genes and genes with median expression below 1. The UMAP is over all SST CALB1 neurons (Fig. 4D legend) with the DE genes as variable genes; the Methods say it follows the procedure of the integrated space (scaling, PCA with 30 PCs, 2D UMAP) without integration, while the notebook uses 6 PCs and 5 neighbours. Clusters are not in the Methods; the notebook's Louvain clustering on the UMAP at resolution 1 is used. The Methods' reading (pairwise FindMarkers, 30 PCs) is the main one and the notebook's a variant, as the protocol puts the Methods first. The cells entering the DE test are open, read as the culture_immune cells of the notebook or all SST CALB1 cells with a DBC, MC or sparse SST call; the adjusted p cut is open, read as 0.05 (Seurat's default report). For the 253, reproduced when the count is within tolerance under one reading. For the clusters, reproduced when three clusters form and each of DBC, MC and sparse SST has most of its cells in a different cluster under the main reading; the variant is shown. The notebook warns that this is "somewhat circular", which is reported.
- fig5a-dbc-two-ttypes
Morphologically classic double bouquet cells map to two t-types, SST CALB1 and SST ADGRG6.
We found that these classical DBC morphologies map to two closely related transcriptomic types, SST CALB1, mentioned above, and L3–5 SST ADGRG6
Results, Double bouquet cells, paragraph 1; Figure 5AHow it will be judged
Reproduced when the cells called DBC in dbc_calb1_ADGRG6.csv include cells of both t-types (Tree_first_cl), seven in SST CALB1 and four in SST ADGRG6 (the counts of figs7a-dbc-regions), and the tree mapping of their transcriptomes gives the same t-types.
- fig5b-merfish-layers
In MERFISH, SST CALB1 is mostly restricted to L2 and SST ADGRG6 to L3, in line with the Patch-seq soma positions, and SST CALB1 DBC axon peaks more superficially.
MERFISH analysis to map the spatial distributions of all types in the MTG taxonomy revealed that the SST CALB1 transcriptomic type was mostly restricted to L2, whereas SST ADGRG6 was mostly restricted to L3
Results, Double bouquet cells, paragraph 1; Figure 5A and 5BHow it will be judged
The modal layer of MERFISH cells mapped to each t-type. Reproduced when it is L2 for SST CALB1 and L3 for SST ADGRG6. The Patch-seq soma layers (Cortical_layer) of the two t-types are reported; they can be tested on the labels alone. The axon part is tested on the layer-aligned DBC reconstructions (Brain Image Library), by the depth of the peak of each DBC's axon branch-length histogram (5 um bins, Methods, Morphology feature analysis); reproduced for that part when the SST CALB1 median peak depth is shallower than the SST ADGRG6 median. No test is reported.
- figs7a-dbc-regions
Six of seven DBCs in SST CALB1 come from temporal cortex and three of four DBCs in SST ADGRG6 from frontal cortex.
We also found that six out of seven DBCs in the SST CALB1 transcriptomic type were from the temporal cortex, whereas three out of four DBCs in the SST ADGRG6 transcriptomic type were from the frontal cortex
Results, Double bouquet cells, paragraph 1; fig. S7A (preprint Supplementary Figure 3A)How it will be judged
Counts are exact (tolerance 0, as they are small integers). DBC calls from dbc_calb1_ADGRG6.csv, region from the NWB metadata where present and otherwise from the lobe column of the metadata table. The counts are taken over all DBCs and, as a variant, over those with an NWB file in the dandiset. No test is reported.
- figs7a-dbc-markers
Most DBCs strongly express CALB1, only one expresses CALB2, all but one express some SST, and TAC1, NOS1 and NPY expression is mixed, so no gene specific to DBCs was found.
Most DBCs showed strong expression of CALB1 (calbindin), whereas only one showed expression of CALB2 (calretinin). All but one neuron showed some degree of SST expression and mixed expression of TAC1, NOS1, and NPY
Results, Double bouquet cells, paragraph 2; fig. S7AHow it will be judged
Expression (CPM, detected when > 0) of CALB1, CALB2, SST, TAC1, NOS1 and NPY in the DBCs of dbc_calb1_ADGRG6.csv. "Strong" is read as log2(CPM + 1) above 5, a threshold the paper does not give; the CALB1 values are reported. Reproduced when more than half of DBCs strongly express CALB1, exactly one has CALB2 > 0, all but one have SST > 0, and each of TAC1, NOS1 and NPY is detected in some but not all DBCs.
- fig5c-dbc-higher-sag
A higher sag ratio is a consistent and prominent feature of all DBCs.
One main feature that was consistent and prominent with all DBCs was a higher sag ratio
Results, Double bouquet cells, paragraph 3; Figures 4C and 5CHow it will be judged
The comparison group is not named; readings are (a) non-DBC neurons of SST CALB1 and SST ADGRG6 (Fig. 5C) and (b) MC and sparse SST neurons of SST CALB1 (Fig. 4C). Sag as in fig2g-sag-by-subclass. No test is reported. Reproduced when, under both readings, the DBC median sag is higher and every DBC with electrophysiology has sag above the comparison group's median; partially reproduced when the median is higher but some DBCs fall below. A two-sided Mann-Whitney test is reported.
- fig5d-dbc-ephys-intermingled
In electrophysiology space, DBCs of the SST CALB1 and SST ADGRG6 types are intermingled with other SST neurons.
Furthermore, in the electrophysiological UMAP space, DBCs from the SST CALB1 and SST ADGRG6 transcriptomic types were intermingled with other SST neurons
Results, Double bouquet cells, paragraph 3; Figure 5D, middleHow it will be judged
No UMAP parameters are given, so a measure is fixed here. SST subclass cells, features of reading (a) z-scored; for each DBC, the fraction of its 10 nearest SST neighbours (feature space and default umap-learn embedding) that are DBCs. Reproduced when the mean fraction is no more than twice the DBC share of SST cells in both spaces, that is, DBCs do not form their own cluster.
- fig5e-bouton-density
Bouton density does not differ between DBCs of the two t-types.
Bouton density measurements did not show differences between the two types
Results, Double bouquet cells, paragraph 4; Figure 5E; fig. S7B (preprint Supplementary Figure 3C)How it will be judged
A claim of absence. The test is not named in any text read; the Methods give the Mann-Whitney test for binary comparisons of morphological features, so a two-sided Mann-Whitney test of bouton density, SST CALB1 against SST ADGRG6 DBCs, at p < 0.05. Reproduced when it is not significant. How boutons were counted is not given in the text read.
- fig6b-species-morphology
For homologous types Pvalb 2 and Sst 5, human neurons have less axonal and dendritic branching but a larger spatial extent than mouse neurons, with no difference in total axon length.
Human neurons had less axonal and dendritic branching yet occupied a larger spatial extent. We found no variation in axon total length between the homologous types
Results, Human versus mouse, paragraph 2; Figure 6BHow it will be judged
Homologous types as the Results define them. Human versus mouse two-sided Mann-Whitney per feature, FDR-corrected (Fig. 6B legend), on axon_num_branches, basal_dendrite_num_branches, axon and dendrite extent (extent_x, extent_y, max euclidean distance) and axon_total_length. Reproduced when, in both homologous types, branch numbers are lower and at least one extent feature is higher in human at FDR p < 0.05, and axon total length does not differ (the absence part, same test).
- fig6c-species-ephys-umap
In electrophysiology UMAP space human and mouse neurons separate clearly for both Pvalb 2 and Sst 5.
In the electrophysiological UMAP, both Pvalb 2 and Sst 5 showed clear separation between humans and mice
Results, Human versus mouse, paragraph 3; Figure 6CHow it will be judged
Human cells of the homologous types (as in fig6e-species-sag-rheobase-width), mouse cells of the mouse t-types the Results list, from Gouwens et al. 2020 (ref. 16). Features computed with IPFX on both. No UMAP parameters are given, so a measure is fixed here; for each cell, the fraction of its 15 nearest neighbours (z-scored features, and a default umap-learn embedding) of its own species. Reproduced when the mean is at least 0.9 for both homologous types in both spaces.
- fig6e-species-sag-rheobase-width
Human and mouse neurons of homologous types Pvalb 2 and Sst 5 differ in sag, rheobase and AP half-width.
Average traces from each type and species shown in Fig. 6D and quantified in Fig. 6E highlight the differences in sag, rheobase, and AP half-width
Results, Human versus mouse, paragraph 3; Figure 6D and 6EHow it will be judged
Human Pvalb 2 is PVALB WFDC2 and L4-6 PVALB SULF1, human Sst 5 is SST CALB1 (M_H_Hodge_cross_species_types in the metadata as a check); mouse Pvalb 2 is Pvalb Calb1 Sst, Pvalb Reln Itm2a, Pvalb Reln Tac1 and Pvalb Sema3e Kank4, mouse Sst 5 is Sst Calb2 Necab1, Sst Calb2 Pdlim5, Sst Mme Fam114a1, Sst Tac1 Htr1d and Sst Tac1 Tacr3 (ref. 16). The Results restrict to supragranular neurons; read as (a) somas in L2 or L3 in human and L2/3 in mouse and (b) all layers, the layer of mouse cells as ref. 16 gives it. Human acute and culture cells are pooled, as the main analysis does (fig. S9 separates them). Two-sided Mann-Whitney human against mouse, FDR-corrected across the features tested. Reproduced when sag, rheobase_i and AP width (width_rheo; width_short_square as a second reading) each differ at FDR p < 0.0001 in both types, under every reading; the directions are reported, as the Results do not state them.
- figs8b-firing-differs-pvalb2-only
Average firing rate and f-I slope differ between species for Pvalb 2 but not for Sst 5.
Features related to AP firing such as average firing rate and AP frequency-current curve slope were found to differ across species for the Pvalb 2 but not for the Sst 5 homologous type
Results, Human versus mouse, paragraph 3; Figure 6D; fig. S8BHow it will be judged
Cells and readings as in fig6e-species-sag-rheobase-width; avg_rate (rheobase + 40 pA sweep as first reading, rheobase sweep as second) and fi_fit_slope. Two-sided FDR-corrected Mann-Whitney (Fig. 6 legend). Reproduced when both differ at FDR p < 0.05 for Pvalb 2, and neither does for Sst 5 (the absence part, judged on the same test), under every reading.
- figs9-species-differences-not-culture
Only the Pvalb 2 species difference in AP width is driven mainly by the culture condition; the other species differences in electrophysiology hold for acute and for culture human neurons alike.
Only the species difference in AP width for Pvalb 2 was driven primarily by the culture condition, whereas all other species differences in electrophysiological features were robustly detected irrespective of acute versus culture paradigm
Results, Human versus mouse, paragraph 4; fig. S9How it will be judged
The fig. S9 legend is missing, so the test is read from the Results. For each feature that differs between species in fig6e-species-sag-rheobase-width and figs8b-firing-differs-pvalb2-only, human acute cells against mouse and human culture cells against mouse, each by two-sided Mann-Whitney, FDR-corrected across features. Reproduced when every such feature except Pvalb 2 AP width differs at FDR p < 0.05 in the same direction in both comparisons, and Pvalb 2 AP width differs for culture but not for acute human cells.
- figs5-vip-enhancer-enrichment
Neurons labelled with the VIP enhancer AAV CN2039 did not map only to VIP types but were strongly enriched for VIP types.
Although the viral-labeled neurons recorded by Patch-seq did not exclusively map to VIP types, we found strong enrichment for diverse VIP neuron types with small cell bodies and bipolar dendrites in L2 and L3
Results, Comparison of features across human GABAergic subclasses, paragraph 2; fig. S5How it will be judged
Cells with Virus CN2039 against other cells, by Revised_subclass_label. The comparison group is not named; readings are (a) all cells without CN2039 and (b) cells labelled with CN1390 only. No test is reported, so "strong enrichment" is judged by a two-sided Fisher exact test, a choice. Reproduced when not all CN2039 cells are VIP, and the VIP fraction among CN2039 cells exceeds that among the comparison cells at Fisher p < 0.05, under both readings. Whether more than half of CN2039 cells are VIP is reported; the paper does not say.
- figs1a-genes-across-days
The higher number of genes detected in culture is present at every day in culture sampled and stays consistent across days in culture.
This elevation was prevalent and remained consistent across all days sampled in culture
Results, Morphological, electrophysiological, and transcriptomic features in the slice culture paradigm, paragraph 1; fig. S1A (preprint Supplementary Figure 1A, genes against days in culture)How it will be judged
No test is reported (the fig. S1 legend is missing; the preprint shows a scatter plot). The consistency part is a claim of absence, judged on the test the Methods give for days in culture, Pearson's correlation of genes detected with Days_In_culture over culture cells; it holds when p >= 0.05. The elevation part holds when the median genes detected at every day in culture sampled exceeds the acute median. Reproduced when both hold; the correlation and the per-day medians are reported.
- fig1e-wfdc2-transcriptome-overlap
Acute and culture PVALB WFDC2 neurons overlap in transcriptomic UMAP space.
For this transcriptomic type, the acute and culture datasets exhibited overlap in transcriptomic UMAP space
Results, Morphological, electrophysiological, and transcriptomic features in the slice culture paradigm, paragraph 2; Figure 1EHow it will be judged
The space is the integrated transcriptomic space of the Methods (Visualization of integrated transcriptomic space; four groups, anchors on the tree-mapping markers, 30 PCs), restricted to the PVALB subclass (Fig. 1E legend). The paper gives no overlap measure, so the measure of fig1i-wfdc2-ephys-umap-overlap is used, on the 30 PCs and on the 2D UMAP. Reproduced when it holds in both.
- fig1h-wfdc2-morph-umap-overlap
In morphological UMAP space, culture and acute PVALB WFDC2 neurons occupy similar space in the PVALB-rich portion.
dimensionality reduction of the morphological features showed that PVALB WFDC2 neurons from the culture and acute paradigms occupied similar space in the PVALB-rich portion of morphological UMAP
Results, Morphological, electrophysiological, and transcriptomic features in the slice culture paradigm, paragraph 2; Figure 1HHow it will be judged
No UMAP parameters or overlap measure are given, so the measure of fig1i-wfdc2-ephys-umap-overlap is used on the morphology features of all reconstructed neurons (z-scored, missing values set to the median), with 5 neighbours as the groups are small (9 and 10 depth-matched cells, or all PVALB WFDC2 reconstructions as a second reading). The PVALB-rich part holds when most of each PVALB WFDC2 cell's 5 nearest neighbours among all cells are PVALB subclass cells, for most cells of each group. Both neighbour count and thresholds are choices. Reproduced when both parts hold in feature space and in a default umap-learn embedding.
- figs1de-mapping-unaffected
Microglial signatures do not impair mapping, as neurons of the same tree-mapped subclass colocalize in transcriptomic UMAP space, whether integrating by paradigm and microglial signature or using marker genes only.
Microglial gene signatures did not impact the ability to accurately map transcriptomic types, as neurons of the same subclass defined from tree mapping colocalized when visualizing the transcriptome through a uniform manifold approximation projection (UMAP)
Results, Morphological, electrophysiological, and transcriptomic features in the slice culture paradigm, paragraph 1; fig. S1D and S1E (preprint Supplementary Figure 1D)How it will be judged
Two spaces, the integrated space of the Methods (Visualization of integrated transcriptomic space) and a space built on the tree-mapping marker genes only without integration (the notebook's chunk "no data integration", read at 5eb357a). No measure is given, so the same-subclass neighbour measure of fig2b-ephys-umap-subclasses is used, within each of the four groups (acute, acute microglial, culture, culture microglial). Reproduced when every subclass's mean same-subclass fraction exceeds its share of cells by at least 0.2 in every group and both spaces; thresholds as in fig2b-ephys-umap-subclasses, choices.
- fig2a-patchseq-overlaps-snrnaseq
In Seurat-aligned transcriptomic UMAP space, the Patch-seq GABAergic neurons overlap with the human MTG snRNA-seq data.
When visualized in Seurat-aligned transcriptomic space through UMAP, the Patch-seq dataset overlapped with single-nucleus RNA sequencing (snRNA-seq) human MTG dataset
Results, Comparison of features across human GABAergic subclasses, paragraph 1; Figure 2AHow it will be judged
The integrated space of the Methods (Visualization of integrated transcriptomic space). No overlap measure is given, so one is fixed here, a choice. For each Patch-seq cell, the subclass held by most of its 15 nearest snRNA-seq nuclei (30 PCs). Reproduced when this equals the tree-mapped subclass for at least 80% of Patch-seq cells; the fraction is reported by subclass.
- fig5d-dbc-morph-umap
In morphological UMAP space, DBCs of the SST CALB1 and SST ADGRG6 types cluster near each other.
In morphological UMAP space, we saw DBCs from these two transcriptomic types cluster near each other with perhaps some separation at the transcriptomic-type level
Results, Double bouquet cells, paragraph 3; Figure 5D, rightHow it will be judged
No UMAP parameters are given, so a measure is fixed here, a choice. SST subclass reconstructions, morphology features z-scored; for each DBC, the fraction of its 5 nearest SST neighbours (feature space and a default umap-learn embedding) that are DBCs. Reproduced when the mean fraction is at least twice the DBC share of SST reconstructions in both spaces. The hedged separation by t-type is not judged; the fraction of each DBC's DBC neighbours of its own t-type is reported.
Pipeline
- Paper
- done (1 of 1)
- Claims
- done
- Claims check
- done (ready)
- Attempt
- todo
- Questions
- none
- Verified
- none