Claims 32
From the paper Cell Atlas and Novel Signal Processing Strategy in Primate Insular Cortex · claims file · checked 2026-10-21: ready
- patchseq-cell-table-link
The dandiset's recordings are the Patch-seq cells of the paper's code table metaInfo_ephysio_patchSeq.csv, with the same subjects, expert cell types and predicted transcriptomic clusters, and the features computed from them agree with the paper's per-cell feature table.
https://dandiarchive.org/dandiset/001746 (main dataset, Patch-seq)
Data availability; github.com/RFLiu2021/AIC_proj at commit 44cd8a6, ephysio/data_ephy/rawdata_ephysio/metaInfo_ephysio_patchSeq.csv and ephysio/data_ephy/Ins_patchseq_ephs.xlsxHow it will be judged
Recordings are matched to table rows by NWB file name (filename_nwb), and through the cell ID (cell_ID) to the rows of Ins_patchseq_ephs.xlsx. Where the dataset carries a subject, cell type or transcriptomic type, it is compared with the table's. For each matched cell in the xlsx, AP half-width, AP amplitude and maximal firing rate are computed as the Methods define them (Extraction of electrophysiological features; first AP of the rheobase sweep, maximal number of APs over 600 ms). The link holds (reproduced) when at least 95% of the 462 rows are found, every label the dataset gives agrees with the table, and at least 80% of matched cells are within 10% of the xlsx value for each of the three features. Partially reproduced when files and labels match but fewer features agree; not reproduced when files are missing or labels disagree. The six Fig. 2 groups formed by the code's cluster lists (see the opening comment) are counted and set against the legend's 30, 17, 119, 22, 16 and 13; the difference is reported and does not decide.
- patchseq-sample
Of 578 Patch-seq cells profiled across all layers of the AIC, 462 were characterized for their intrinsic membrane properties; the Patch-seq experiments used 12 male macaques.
profiled 578 cells across all layers of the macaque AIC. A total of 462 neurons were characterized in terms of their intrinsic membrane properties ... Patch-seq ( n = 12, all males)
Results, Patch-seq revealed extensive cellular heterogeneity in the primate AIC, paragraph 1; Methods, AnimalsHow it will be judged
Neurons are the recorded cells with a current-clamp step protocol; animals are the distinct subjects. Reproduced when both counts are within tolerance and every subject is male; partially reproduced when the neuron count holds but the animal count or sex does not.
- patchseq-mapping
Of the Patch-seq cells with transcriptomes, 307 were analysed; 273 excitatory neurons map to 18 of the 20 excitatory types and 34 inhibitory neurons to 17 of the 33 inhibitory types.
Among them, 307 cells had their transcriptomes recovered and were used for downstream analysis ... 273 excitatory neurons mapped to 18 of the 20 excitatory types, while 34 inhibitory neurons mapped to 17 of the 33 inhibitory types
Results, Patch-seq revealed extensive cellular heterogeneity in the primate AIC, paragraph 1; Figure 1e; Extended Data Figure 6aHow it will be judged
Counted from the transcriptomic type of each recording (the dataset's metadata, or predictCluster_label of metaInfo_ephysio_patchSeq.csv), Exc types against Inh types. Cells whose mapping rate is below 30% are excluded (Methods, Mapping of Patch-seq cells). The mapping itself (GEO GSE319369 reads) is not redone. Reproduced when every count is within tolerance.
- fig2a-subclass-counts
The 273 excitatory Patch-seq neurons fall in seven of the eight subclasses, L2/3 IT HPCAL1 (33), IT RORB (23), L5 IT FEZF2 (143), L5 ET (20), L5/6 CT (11), L6 IT (5) and L6b (9).
The 273 Patch-seq excitatory neurons were assigned into seven of the eight major subclasses (Fig. 2a,b ): L2/3 IT HPCAL1 ( n = 33), IT RORB ( n = 23), L5 IT FEZF2 ( n = 143), L5 ET ( n = 20), L5/6 CT ( n = 11), L6 IT ( n = 5) and L6b ( n = 9)
Results, Patch-seq revealed extensive cellular heterogeneity in the primate AIC, paragraph 2; Figure 2a,bHow it will be judged
Subclass of each excitatory cell from its transcriptomic type. The paper names the types of each subclass only through its code. Readings, (a) the code's cluster lists of the opening comment; (b) the subclass the dataset's metadata gives, where it gives one. The seven stated counts sum to 244, and 244 plus the 29 VENs of ven-map-et-ct is 273, so the stated counts leave the VENs out; under both readings the VENs (expert type VEN) are counted apart and not in L5 ET or L5/6 CT. The code merges L6 IT and L6b into L6 PC (clusters 4, 12 and 17; the Results give L6 PC n = 14); the Methods' rule (HPCAL1 clusters in L6 are L6b, OPRK1 clusters L6 IT) puts Exc HPCAL1 KANK4 in L6b, and the marker of clusters 4 and 12 is not given, so under (a) L6 IT and L6b are judged together against 5 + 9 = 14. The counts with the VENs included are shown as a variant that does not decide. Reproduced when every count is within tolerance under one reading; partially reproduced when at least five of the seven are.
- ven-map-et-ct
The 29 morphologically identified VENs do not map to one transcriptomic cluster; 14 map to L5 ET and 15 to L5/6 CT.
29 neurons that were morphologically identified as VENs did not map to a unique molecular cluster; instead, they mapped to either L5 ET ( n = 14) or L5/6 CT ( n = 15)
Results, Patch-seq revealed extensive cellular heterogeneity in the primate AIC, paragraph 2; Figure 2aHow it will be judged
VENs are recordings whose expert type is VEN and that have a transcriptomic type (Patch-seq cells with mapping rate of 30% or more). Reproduced when both counts are within tolerance and no such VEN maps to another cluster; partially reproduced when the counts hold but other clusters also hold VENs (their number is reported).
- fig6f-ven-subtype-mapping
VENs mapped to L5 ET (Exc FEZF2 DSG2) are VEN-Ls (n = 10) and those mapped to L5/6 CT (Exc FEZF2 COL24A1) are VEN-Ss (n = 11).
the morphologically defined VENs mapped to L5 ET (Exc FEZF2 DSG2) were VEN-Ls ( n = 10), while those VENs mapped to L5/6 CT (Exc FEZF2 COL24A1) were VEN-Ss ( n = 11)
Results, Multimodal profiling of two VEN subtypes, paragraph 2; Figure 6fHow it will be judged
VEN subtype from the expert label (VEN_Long, VEN_Short) of the dataset or of metaInfo_ephysio_patchSeq.csv, crossed with the transcriptomic type. Reproduced when both counts are within tolerance and no VEN-L maps to Exc FEZF2 COL24A1 or VEN-S to Exc FEZF2 DSG2. The 14 and 15 VENs of ven-map-et-ct exceed 10 and 11; the paper does not say whether the rest lack a traced morphology or a subtype label, and the cross-table is reported.
- fig2d-l23it-small-tau
Among the six projection neuron groups, L2/3 IT HPCAL1 neurons stand out with a small membrane time constant.
L2/3 IT HPCAL1 neurons stood out with a small membrane time constant (tau) and narrow action potential (AP) width (Fig. 2d–f )
Results, Patch-seq revealed extensive cellular heterogeneity in the primate AIC, paragraph 3; Figure 2dHow it will be judged
Groups as in the code (opening comment); the legend counts 30 L2/3 IT HPCAL1, 17 IT RORB, 119 L5 IT FEZF2, 22 VEN, 16 L5 ET and 13 L6 PC neurons. Tau as in the Methods, an exponential f(t) = a exp(-t/tau) + C fitted from step onset to the sag minimum, averaged over all hyperpolarizing traces; the code's first-sweep tau is a variant. Test, two-tailed Wilcoxon rank-sum with Benjamini-Hochberg FDR over the features of the comparison, as calFDR_ephsio.m does (the legend names the test; the paper's FDR values are in Supplementary Table 2, not read). The Methods do not list every feature ("The adaptation indices ... For instance"), so the code settles the FDR family, the 39 features of Ins_patchseq_ephs.xlsx that calFDR_ephsio.m takes, those the Methods define computed as defined there and the others as createFeatures.m computes them. The family of the 20 features the Methods name (rheobase, threshold, amplitude, rise and decay time, half-width, latency, AHP, maximal firing rate, sag value, sag ratio, sag time, sag area, RMP, input resistance, tau, rebound, ISI and amplitude adaptation indices and the latency slope) is a variant. "Stood out" allows two readings, as the code computes both, (a) L2/3 IT HPCAL1 against the other five pooled, and (b) against each other group. Reproduced when its median tau is the lowest of the six groups and the comparison has FDR < 0.05 (conventional) under (a) or every comparison does under (b); partially reproduced when the median is lowest but neither reading is significant. The code's plot exclusion (tau of 1 ms or less) is a variant.
- fig2f-l23it-narrow-ap
L2/3 IT HPCAL1 neurons have a narrow AP width compared with the other projection neuron groups.
L2/3 IT HPCAL1 neurons stood out with a small membrane time constant (tau) and narrow action potential (AP) width (Fig. 2d–f )
Results, Patch-seq revealed extensive cellular heterogeneity in the primate AIC, paragraph 3; Figure 2fHow it will be judged
AP width is the half-width of the first AP of the rheobase sweep, at half the threshold-to-peak amplitude, threshold at the maximum third derivative in the 20 ms before the peak (Methods; the code plots Half_Width_ms as "AP width"). Tests and readings (a) and (b) as in fig2d-l23it-small-tau. Reproduced when its median half-width is the lowest of the six groups and the comparison is significant at FDR < 0.05 (conventional) under (a), or every comparison is under (b); partially reproduced when the median is lowest but neither reading is significant. The code table gives a lower median for IT RORB than for L2/3 IT HPCAL1, so "lowest" may fail with the claim still holding against the pooled groups; the rank and both readings are reported. The code's plot exclusion (widths outside 1 to 5 ms) is a variant.
- supp11-rm-itrorb-l6pc
IT RORB and L6 PC neurons have a distinctive input resistance among projection neuron groups.
Other types exhibit specific characteristics including distinctive membrane resistance in IT RORB and L6 PC neurons
Results, Patch-seq revealed extensive cellular heterogeneity in the primate AIC, paragraph 3; Figure 2e; Supplementary Figure 11How it will be judged
Input resistance as in the Methods, the slope of steady-state voltage (last 100 ms of the step) minus RMP against injected current over the steps without APs; the code takes HEKA sweeps 1 to 6 (-200 to -100 pA) and the last two thirds of the step, a variant. The text does not give the direction; "distinctive" is read as differing from the other groups, and the direction is reported. Test and readings (a), each against the other five pooled, and (b), against each other group, as in fig2d-l23it-small-tau (Supplementary Fig. 11 and Table 2 not read). Reproduced when both IT RORB and L6 PC differ at FDR < 0.05 (conventional) under one reading; partially reproduced when one of them does.
- supp11-l5it-threshold-latency
L5 IT FEZF2 neurons have a higher AP threshold and a longer spike latency than the other projection neuron groups.
Other types exhibit specific characteristics including ... a higher AP threshold and longer spike latency in L5 IT FEZF2 neurons
Results, Patch-seq revealed extensive cellular heterogeneity in the primate AIC, paragraph 3; Supplementary Figure 11How it will be judged
Threshold as in fig2f-l23it-narrow-ap; latency from step onset to the AP threshold of the first AP (Methods), in the rheobase sweep as the code takes it. Test and readings (a) and (b) as in fig2d-l23it-small-tau (Supplementary Fig. 11 not read). Reproduced when, under one reading, L5 IT FEZF2 has the higher median and FDR < 0.05 (conventional) for both features; partially reproduced when for one.
- supp11-ven-broad-ap
VENs have a broader AP than the other Patch-seq projection neuron groups.
Other types exhibit specific characteristics including ... a broader AP width in VENs
Results, Patch-seq revealed extensive cellular heterogeneity in the primate AIC, paragraph 3; Figure 2f; Supplementary Figure 11How it will be judged
Half-width as in fig2f-l23it-narrow-ap. Test and readings (a) and (b) as in fig2d-l23it-small-tau. Reproduced when, under one reading, VENs have the highest median and FDR < 0.05 (conventional); partially reproduced when the median is highest without significance.
- supp11-l5et-deep-sag
L5 ET neurons have a deeper sag than the other projection neuron groups.
Other types exhibit specific characteristics including ... and a deeper sag in L5 ET neurons
Results, Patch-seq revealed extensive cellular heterogeneity in the primate AIC, paragraph 3; Supplementary Figure 11How it will be judged
The Methods define sag value (RMP minus steady state at the largest hyperpolarizing step), sag ratio, sag time and sag area; "deeper" names none, and the code table holds Sag_Ratio. Readings of the measure, (i) sag ratio and (ii) sag value. Test and readings (a) and (b) of the comparison as in fig2d-l23it-small-tau. Reproduced when, under every measure reading and one comparison reading, L5 ET has the deepest sag with FDR < 0.05 (conventional); blocked as methods_underspecified when the measure readings disagree.
- ven-firing-rate-variability
VENs have lower firing rates and greater variability in spike timing than PCs.
Notably, VENs exhibit lower firing rate and greater variability in spike timing than PCs
Results, Patch-seq revealed extensive cellular heterogeneity in the primate AIC, paragraph 3; Supplementary Tables 1 and 2How it will be judged
Firing rate is the maximal firing rate, the largest AP count over the 600 ms step (Methods). The text names no measure of spike-timing variability; readings from the code's feature table, (i) ISI CV (std/mean of the ISIs of the maximal-rate sweep, createFeatures.m) and (ii) ISI Fano factor of that sweep. "PCs" is read as (a) the five PC groups pooled, as calFDR_ephsio.m compares VEN against the others, (b) each PC group, and (c) the layer 5 PC groups pooled (L5 IT FEZF2 and L5 ET), as the Discussion says "Compared with neighbouring PCs". Test, two-tailed Wilcoxon rank-sum with Benjamini-Hochberg FDR as in fig2d-l23it-small-tau (Supplementary Tables 1 and 2 not read). Reproduced when, under one comparison reading, VENs have a lower median firing rate and a higher median variability under both measure readings, each at FDR < 0.05 (conventional); partially reproduced when only the firing rate holds; blocked as methods_underspecified when the measure readings disagree.
- l5et-vs-l5it-sag-rm-rate
L5 ET neurons have a deeper sag, a lower input resistance and a higher firing rate than L5 IT neurons.
Specifically, L5 ET neurons in macaque AIC consistently exhibit deeper sag, lower input resistance and higher firing rates than L5 IT neurons.
Discussion, paragraph 1How it will be judged
L5 ET against L5 IT FEZF2, groups as in the code; sag readings as in supp11-l5et-deep-sag, input resistance as in supp11-rm-itrorb-l6pc, maximal firing rate as in ven-firing-rate-variability. Test, two-tailed Wilcoxon rank-sum with Benjamini-Hochberg FDR over the features of the pairwise comparison (calFDR_ephsio.m, FDR_L5IT_vs_L5ET). Reproduced when all three differ in the stated direction at FDR < 0.05 (conventional); partially reproduced when two do.
- fig3d-saddle-phase-plane
The AP phase-plane plots of VENs show a more prominent biphasic shape than those of PCs, with the AIS phase decaying before the SD phase in a saddle-like shape.
VENs showed a more prominent biphasic phase than PCs. Different from PCs, the AIS phase of VENs exhibited a clear decay before the SD phase
Results, The AIC evolved a neural signal-processing strategy, paragraph 3; Figure 3dHow it will be judged
The paper shows six example cells and says it quantifies the 'saddle-like' shape by the phase peak index ("We used an index ... to quantify the 'saddle-like' shape"). The claim is therefore judged by fig3e-phase-peak-index and takes its outcome. The fraction of cells whose phase plot has a local minimum of dV/dt between the AIS and SD peaks (maximal-rate sweep, dV/dt smoothed over 5 samples as in calPhaseHeightIdx.m), VENs against the PC groups pooled, is not a measure the paper uses; it is shown as a description that does not decide.
- fig3e-phase-peak-index
VENs have lower phase peak indices than any PC group.
The VENs had lower phase peak indices than any type of PCs (Fig. 3e )
Results, The AIC evolved a neural signal-processing strategy, paragraph 3; Figure 3eHow it will be judged
Index (Psd - Pais)/(Psd + Pais), Psd and Pais the peaks of the SD and AIS phases of dV/dt against Vm; where a phase has no clear peak, "the point with the lowest derivative value" (Methods, Calculation of the phase peak index). The paper read the peaks by hand (calPhaseHeightIdx.m), so an automatic reading is fixed here and its open parts named. Sweep, the maximal-rate sweep (the code's). Peaks, Pais the first local maximum of smoothed dV/dt (5 samples) on the rising phase and Psd the largest maximum after it; with one maximum only, the point of the rising phase where d2V/dt2 is lowest (the shoulder) before the global maximum stands for the missing peak, which is how "lowest derivative" is read. Open, which APs, readings (a) the first AP of the sweep and (b) the median index over its APs. Test, Kruskal-Wallis over the six groups followed by Dunn's test (legend), with the correction not stated; readings Bonferroni and Sidak, at 0.05 (conventional). Reproduced when, under every AP reading and both corrections, the VEN median is the lowest and Dunn's test separates VENs from every PC group; partially reproduced when the median is lowest and some groups are separated; blocked as methods_underspecified when the AP or correction readings disagree. The automatic peak reading is not the paper's (the peaks were clicked by hand); when its outcome differs from the one the code's per-cell indices give, that is reported as a deviation. The code's per-cell values (PeakIdx_Fig5e.xlsx) are compared with the automatic ones as a check that does not decide.
- fig4ac-ven-broader-smaller-ap
APs of VENs (n = 281) are broader and lower in amplitude than those of L5 PCs (n = 306), throughout the spike train evoked by supra-threshold current.
APs in VENs were broader and had a lower amplitude compared with L5 PCs (Fig. 4a-c ). These differences were consistently observed throughout the spike train evoked by supra-threshold current injections.
Results, The AIC evolved a neural signal-processing strategy, paragraph 4; Figure 4a-cHow it will be judged
The 281 VENs and 306 L5 PCs are "additional" whole-cell recordings; the paper does not say where they are deposited, and the dandiset may hold only the Patch-seq cells. If the dataset holds them, the first near-threshold AP (rheobase sweep) is measured, amplitude and half-width as in the Methods; two-tailed Wilcoxon rank-sum, VEN against PC (Fig. 4a legend). Reproduced when VEN half-width is larger and amplitude smaller, each with p < 0.001, and the counts are within tolerance. "Throughout the spike train" is shown in Fig. 4c by the first and the penultimate AP of the supra-threshold train, with no test; the same comparison is run on the penultimate AP of the maximal-rate sweep and reported, and does not decide. If the dataset holds only the Patch-seq cells, the claim is blocked as data_missing, and the same comparison on the Patch-seq VENs and L5 PCs (the PC_L5 and L5 ET labels of the cell table) is shown as a variant that does not decide.
- fig4i-peak-index-ais-distance-real
In real neurons the phase peak index decreases linearly with the distance of the AIS from the soma (n = 62).
Bottom: real neurons ( n = 62, each data point was averaged of 2–30 neurons except for the rightmost one with n = 1). Data are shown as mean ± s.e.m. Two-tailed Student’s t -test, P < 0.0001 for slope.
Figure 4i legend; Results, The AIC evolved a neural signal-processing strategy, paragraph 5How it will be judged
AIS distance per neuron is a morphological measure the dataset does not hold; it comes from the reconstructions (NeuroMorpho.org doi:10.13021/y4be-0p18) or the code's NeuronPeakIdx_summary.mat (axonDist, 62 neurons), matched to the recordings by cell ID. Which 62 neurons is not said; they are taken as the Patch-seq cells with a reconstructed axon origin. Peak index as in fig3e-phase-peak-index, under its readings. As in Fig4itop_plot_PhasePeakIdx_vs_AISd.m, indices are averaged in 10 um bins of AIS distance and a line fitted to the bin means against the bins' left edges; the slope is tested by a two-tailed t-test (legend). Reproduced when the slope is negative with p < 0.05 under every peak-index reading. A regression on the 62 neurons is a variant. The ANCOVA against the model belongs to model-ais-distance-ap-shape and needs the simulations (data elsewhere, see the claims file of DANDI:001750).
- ext9cd-ven-subtypes-ap
AP width and amplitude do not differ between VEN-L (n = 85) and VEN-S (n = 41).
Despite their distinct morphological and transcriptomic profiles, both VEN subtypes exhibited similar physiological properties (Extended Data Fig. 9 and Supplementary Table 8 )
Results, Multimodal profiling of two VEN subtypes, paragraph 2; Extended Data Figure 9c,dHow it will be judged
A claim of absence, judged on the paper's test, two-tailed Wilcoxon rank-sum with Benjamini-Hochberg FDR over the two features (Extended Data Fig. 9c,d legend), at 0.05 (conventional), on the first AP's half-width and amplitude. Reproduced when neither has FDR < 0.05, with the median differences reported; not reproduced when either does. The 85 and 41 cells are more VENs than the Patch-seq table holds (68) and the code identifies them by number only; the paper does not say where they are deposited. If the dataset does not hold them, the claim is blocked as data_missing and the test on the Patch-seq VEN-L and VEN-S is shown as a variant that does not decide.
- supp8-ven-subtypes-similar-features
Across electrophysiological features, VEN-L and VEN-S recorded without RNA extraction do not differ.
Despite their distinct morphological and transcriptomic profiles, both VEN subtypes exhibited similar physiological properties (Extended Data Fig. 9 and Supplementary Table 8 )
Results, Multimodal profiling of two VEN subtypes, paragraph 2; Supplementary Table 8; github.com/RFLiu2021/AIC_proj at commit 44cd8a6, ephysio/calFDR_VENsubtype_ephysio.mHow it will be judged
Supplementary Table 8 (not read) gives "Quantified features and FDR-adjusted P values for comparisons of electrophysiological features between VEN-L and VEN-S", and the Extended Data Fig. 9 legend says these comparisons were "obtained without RNA extraction", so the paper's cells are the additional whole-cell recordings (as for ext9cd-ven-subtypes-ap), not the Patch-seq cells; the Patch-seq VEN-L against VEN-S comparison belongs to Supplementary Table 9 ("among Patch-seq recorded VEN-L, VEN-S and L5/6 CT"). Where the paper and the code disagree the paper comes first. A claim of absence, judged on the paper's test as in the Extended Data Fig. 9 legend and the code (calFDR_VENsubtype_ephysio.m), two-tailed Wilcoxon rank-sum per feature with Benjamini-Hochberg FDR over the feature family of fig2d-l23it-small-tau, at 0.05 (conventional). If the dataset holds the VEN-L and VEN-S recorded without RNA, with their subtype, reproduced when no feature has FDR < 0.05 and not reproduced when one does. If it does not, the claim is blocked as data_missing, and the same test on the Patch-seq VENs, grouped as calFDR_VENsubtype_ephysio.m does (VEN in cluster 16 against VEN in cluster 8) and by the expert labels VEN_Long and VEN_Short, is shown as a variant that does not decide, with the features of lowest FDR.
- supp9-venl-more-distinct-from-ct
VEN-L differs more from L5/6 CT neurons in physiology than VEN-S does.
Mirroring the transcriptomic similarity between VEN-S and L5/6 CT, VEN-L shows greater physiological disparity from L5/6 CT than does VEN-S (Supplementary Table 9 )
Results, Multimodal profiling of two VEN subtypes, paragraph 2; Supplementary Table 9How it will be judged
As in calFDR_ephysCmp_VEN_vs_L56CT.m, VEN-L (cluster 16 VEN), VEN-S (cluster 8 VEN) and L5/6 CT (cluster 8 PC; n = 10 in the Extended Data Fig. 9i legend); two-tailed Wilcoxon rank-sum per feature with Benjamini-Hochberg FDR over the features of fig2d-l23it-small-tau. "Greater disparity" is not quantified in the text; it is read as more features with FDR < 0.05 (conventional) for VEN-L against L5/6 CT than for VEN-S against L5/6 CT. Reproduced when the count is larger for VEN-L; not reproduced when it is equal or smaller. The median FDR over features is reported as a second measure that does not decide.
- morph-more-distinct-than-ephys
Morphological differences between projection neuron groups are more pronounced than their physiological differences, with more features differing.
Overall, morphological differences between projection neuron types were more pronounced than their physiological differences ( P < 0.05 for more features; Supplementary Table 3 and 4 )
Results, Patch-seq revealed extensive cellular heterogeneity in the primate AIC, paragraph 3; Supplementary Tables 3 and 4, as cited (Supplementary Table 2 holds the electrophysiological FDR values)How it will be judged
Morphological features (40, Methods, Extraction of morphological features) need the reconstructions (NeuroMorpho.org doi:10.13021/y4be-0p18), matched by cell ID. The comparisons are those of calFDR_ephsio.m and its morphological counterpart, each group against the rest and each pair, with BH FDR within each comparison. "More features" can be read as (a) the count or (b) the fraction of features with FDR < 0.05, summed over comparisons; the feature counts differ (40 morphological, the Methods' electrophysiological features). Reproduced when morphology exceeds physiology under both readings; blocked as methods_underspecified when they disagree.
- fig7c-tau-rna-distance
Across pairs of projection neuron types, distances in membrane time constant correlate with transcriptomic distances (r = 0.53), and other morphoelectric properties show similar correlations.
Pooling the pairs across all clusters revealed a high correlation ( r = 0.53) between the membrane time constant (tau) distances and RNA distances (Fig. 7a–c ). Similar correlations extended to other morphoelectric properties
Results, Transcriptomic distinctions underlie phenotypic variation, paragraph 1; Figure 7a-iHow it will be judged
Tau per cell from the dataset (fig2d-l23it-small-tau, both of its tau readings); the RNA side needs the scRNA-seq counts (GEO GSE319557). Per transcriptomic type, the tau distance is the Euclidean distance between type means; the transcriptomic distance is the correlation distance (1 - Pearson r) of mean log2(x + 1) 10x counts over the 500 most variable genes (Methods, Variability analysis; Results, "correlation distance"). Types, the 12 clusters with more than five mapped Patch-seq neurons, as the code's notebook (Fig7cf_supplFig16a_ephyFeatures_2_10X_Ext.ipynb) lists them (0, 1, 2, 4, 5, 6, 8, 14, 15, 16, 17, 18); the Methods give no minimum. Pearson r over all pairs. The code instead takes the log of the mean Patch-seq counts over its own variable genes; that is shown as a variant. Reproduced when r is within tolerance with the Methods' tau; partially reproduced when r is positive with p < 0.05 but outside tolerance. The code's first-sweep tau is a variant that does not decide. The maximal firing rate (Fig. 7d-f) is shown the same way, with no stated r, and does not decide.
- scrna-classes
The whole-cell scRNA-seq atlas of the macaque AIC holds 122,663 transcriptomes in four classes, excitatory neurons (21,733, 17.8%), inhibitory neurons (17,007, 13.9%), glial cells (83,283, 67.8%) and vascular cells (640, 0.5%), divided into 78 transcriptomic types.
we categorized these cells into four main classes: excitatory neurons (21,733 cells, 17.8%), inhibitory neurons (17,007 cells, 13.9%), glial cells (83,283 cells, 67.8%) and vascular cells (640 cells, 0.5%)
Results, Transcriptomic landscape of the macaque AIC, paragraph 2; Figure 1b,c; Extended Data Figure 1e,fHow it will be judged
Counted from the deposited cluster labels; tested in full under the claims file of DANDI:001750 (scrna-classes, scrna-78-types).
- ext1d-ven-soma-density
VENs are found only in layer 5b of the AIC and ACC; Extended Data Fig. 1d compares soma size and neuron density between AIC and V1 in layers 2/3 and 5, with no result stated in the text read.
The VENs, characterized by spindle-shaped perikarya and bipolar thick dendrites, were exclusively identified in layer 5b of the macaque AIC and anterior cingulate cortex (ACC)
Results, Transcriptomic landscape of the macaque AIC, paragraph 1; Extended Data Figure 1a-dHow it will be judged
Not testable on this dataset. The text does not state the direction of the soma size and density differences; judged from the Source Data, two-sided Wilcoxon rank-sum per layer, AIC against V1, with the correction under both readings the legend names (Bonferroni and Benjamini-Hochberg), at 0.05 (conventional), and the direction reported. Soma size from 105 AIC L2/3, 85 V1 L2/3, 90 AIC L5 and 90 V1 L5 cells; density from 15, 20, 15 and 20 ROIs, three animals per region (legend).
- fig3c-ais-distance-length
In Pan-Nav stained sections the AIS of VENs lies farther from the soma than that of PCs, and VENs have a shorter AIS.
showed that the AIS in VENs was, on average, located significantly farther from the soma than in PCs (Fig. 3c ). Additionally, VENs had a shorter AIS than PCs.
Results, The AIC evolved a neural signal-processing strategy, paragraph 2; Figure 3cHow it will be judged
Tested in full under the claims file of DANDI:001750 (fig3c-ais-distance-length), with the readings of the ANOVA's factors named there.
- fig4d-sodium-current-lower
Somatic Na+ currents in outside-out patches are lower in VENs (n = 14) than in L5 PCs (n = 21).
outside-out patch-clamp recording showed that the somatic Na + currents evoked in VENs ( n = 14) were significantly lower than those of L5 PCs ( n = 21) (Fig. 4d )
Results, The AIC evolved a neural signal-processing strategy, paragraph 4; Figure 4dHow it will be judged
Tested in full under the claims file of DANDI:001750.
- model-ais-distance-ap-shape
In the VEN model, moving the AIS farther from the soma makes APs wider and smaller and the phase-plane plot more saddle-like, while the PC model does not show this; the models reproduce the measured differences in AP amplitude and width, and the VEN model's linear decrease of the phase peak index with AIS distance matches that of the real neurons.
Crucially, simulating an increase in the distance of the AIS from the soma in the VEN model led to APs being wider and smaller in amplitude, and the phase–plane plot becoming more ‘saddle-like’, (Fig. 4g and Extended Data Fig. 7b ). This effect was not replicated in the PC model (Fig. 4h ). Furthermore, the linear decrease in the phase peak index with increasing AIS distance observed in the VEN model closely matched the experimental data
Results, The AIC evolved a neural signal-processing strategy, paragraph 5; Figure 4e-i; Extended Data Figure 7bHow it will be judged
Tested under the claims file of DANDI:001750 (fig4g-model-ais-distance-ap-shape and fig4i-peak-index-ais-distance), with the criteria named there.
- fig4j-eepsp-ven-larger
Paired VENs and PCs respond to deep-layer stimulation with similar latencies (5.13 and 5.02 ms), and VEN eEPSPs are larger, for VEN-L and VEN-S alike.
The postsynaptic response latencies were similar between VENs and PCs (5.13 ± 0.29 ms for VENs and 5.02 ± 0.30 ms for PCs, two-tailed paired t -test ... whereas the response amplitudes in VENs were significantly higher than those in adjacent PCs
Results, VENs respond robustly to deep-layer synaptic input, paragraph 1; Figure 4j; Extended Data Figure 9e,fHow it will be judged
Tested under the claims file of DANDI:001750 (fig4j-eepsp-latency and fig4j-eepsp-amplitude), with the readings named there.
- fig5-psc-and-connectivity
VENs receive fewer spontaneous and miniature excitatory and inhibitory events than PCs at similar amplitudes, and form almost no local monosynaptic connections (PC-to-PC 24 of 503 pairs; VEN-to-VEN 0 of 110, VEN-to-PC 0 of 200, PC-to-VEN 1 of 200).
The frequencies of both excitatory and inhibitory events were significantly lower in VENs than in PCs (Fig. 5b ), whereas the amplitudes were not significantly different between VENs and PCs (Fig. 5c ) ... PCs formed monosynaptic connections with other PCs in 24 out of 503 tested pairs
Results, Altered connectivity profile of deep-layer AIC neurons, paragraphs 1 and 2; Figure 5b-iHow it will be judged
Tested in full under the claims files of DANDI:001752 and DANDI:001751.
- fig6-ven-subtype-morphology-markers
VEN-L has a long prominent basal dendrite and a larger basal arborization than PCs, VEN-S a short one similar to PCs; both have the axon origin farther from the soma than PCs. VEN-L expresses DSG2, HAPLN4, TOX2, TRPC4, CAMK1G and TAGLN2, VEN-S ATP8A2, POC5 and SEMA3E.
VEN-L had larger basal dendritic arborization than L5 PCs, whereas VEN-S had intermediate-sized basal dendrite arborization similar to PCs ... VEN-L expressed markers such as DSG2, HAPLN4, TOX2, TRPC4, CAMK1G and TAGLN2 , whereas VEN-S showed high expression of ATP8A2, POC5 and SEMA3E
Results, Multimodal profiling of two VEN subtypes, paragraphs 1 and 2; Figure 6a-jHow it will be judged
Tested under the claims file of DANDI:001750 (fig6-ven-subtype-morphology and fig6h-subtype-markers), with the readings named there.
- fig8-metabolomics
Of 581 metabolites, 228 differ between AIC and V1 and 108 between AIC and parietal cortex, and AIC shows a glycolytic bias with elevated citrate, reduced 2-ketobutyrate and increased glutamate.
we identified 228 differentially abundant metabolites between the AIC and V1 (Welch’s ANOVA test followed by Games–Howell test, P < 0.05) compared with only 108 between the AIC and PrC
Results, Cortical metabolomics reveals a glycolytic bias of the AIC, paragraphs 1 to 4; Figure 8How it will be judged
Tested under the claims file of DANDI:001750 (fig8a-metabolite-gradient and fig8de-glycolytic-bias).
Pipeline
- Paper
- done (1 of 1)
- Claims
- done
- Claims check
- done (ready)
- Attempt
- todo
- Questions
- none
- Verified
- none