Figure 2 - available via license: Creative Commons Attribution 4.0 International
Content may be subject to copyright.
Ephys response genes are subclass-and ephys feature-specific. (a) Stacked bar chart indicating, for each Sst ephys feature, the fraction of the corresponding correlation-ephys genes that are also correlation-ephys genes for other ephys features in Sst. Correlation-ephys genes are grouped based on how many other ephys features they are correlated with. Features are sorted based on average number of other ephys features that each ephys gene is correlated with. (b) Stacked bar chart indicating, for each ephys feature, the percentage of all genes correlated with that feature (taken over all four subclasses) that are shared by either one (unique), two or three subclasses. No ephys gene was correlated with the same ephys feature across all four subclasses. (c) Plot of average AP dV/dt, Subthresh. (abs.), and ISI shape as a function of time steps for the genes Cntn5, Col11a1 and Unc13c, respectively. For AP dV/dt, time point 1 references the initiation of the first action potential. For Subthresh. (abs.), values represent the concatenated responses to hyperpolarizing current steps (from -10 pA to -90 pA, in steps of 20pA). For ISI shape, the average of ISI voltage trajectories are visualized, aligned to the threshold of the initial AP and normalized in duration. Each of the three genes are ephys genes specific to Sst, and are not ephys genes of the other three subclasses. Lines indicate the average ephys behavior of neurons at the top (orange) and bottom (blue) 10 th percentile in expression of the three genes for each of the four subclasses.
Source publication
Multimodal assays such as Patch-seq that simultaneously profile molecular and cellular phenotypes of cells enable the identification of molecular underpinnings of electrophysiological response patterns in neurons. Here we analyzed Patch-seq measurements of thousands of mouse interneurons to identify subclass-specific genes associated with different...
Contexts in source publication
Context 1
... established that many ephys features are well predicted by gene expression, we next wondered to what extent is the transcriptional ephys response subclass specific. It is already appreciated that different interneuron types have different ephys properties 52 ( Supplementary Fig. S2), but what has not been explored is whether the same genes underlie the ephys response of different interneuron classes. To test this hypothesis, we first constructed predictive ephys models separately for each of the four interneuron subclasses, and compared model prediction performance. ...
Context 2
... as the percentage of correlation-ephys genes shared with fewer than five ephys features in the entire subclass. On average across the 42 ephys features in Sst neurons for example, 25% of correlation-ephys genes were only correlated to less than five ephys features, indicating a high degree of specificity of genes to specific ephys features (Fig. 2a); other subclasses showed similar levels of specificity (Supplementary Fig. S4). These results indicate different genes contribute to different ephys response features in the same ...
Context 3
... of gene overlap of ephys genes for the same ephys feature across subclasses, we calculated a subclass specificity score measured as the average percentage of correlation-ephys genes specific to each subclass, when considering the same ephys feature across subclasses. We found on average subclass specificity of 90.6% across the 42 ephys features (Fig. 2b), indicating that even for the same ephys feature, different genes establish a response to ephys stimulus. Our results were consistent when we performed this analysis on the genes identified by our predictive models: we found minimal overlap in genes selected by models trained on the same interneuron class but across different ephys ...
Context 4
... visualize the subclass-specific effect of ephys genes, Figure 2c illustrates the first action potential recorded for neurons from each subclass, when averaged for neurons exhibiting either low or high levels of expression of the three ephys genes Cntn5, Cdh7 and Reln predicted to specific to Sst. For each of the three ephys genes, a significantly larger difference in action potential behavior is observed for Sst interneurons. ...
Context 5
... heritability signal of the AP dV/dt feature was specific to Schizophrenia; no other psychiatric disorders or neural disorders showed significant heritability, though Bipolar disorder shows more nominal signal than other traits (Fig. 3b). The specificity of the heritability signal to Schizophrenia does not appear to be driven by statistical power of the underlying GWAS; there were other tested disorders that had substantially larger cohort sizes and number of genomewide significant loci in their GWAS (Supplementary Table 2). These results suggest that action potential voltage changes in Sst ephys responses may play a role in mediating Schizophrenia genetic risk. ...
Context 6
... features in Sst, we found cell adhesion genes were enriched in four ephys feature sets, composed of AP dV/dt, AP width, and up/downstroke ratios (Supplementary Fig. S9). The enrichment of cell adhesion genes across multiple Sst ephys features was surprising, given we saw the poor overlap in associated genes between different ephys features of Sst (Fig. 2a), and poor overlap of ephys features between cell types (Fig. ...
Context 7
... ephys feature sets, composed of AP dV/dt, AP width, and up/downstroke ratios (Supplementary Fig. S9). The enrichment of cell adhesion genes across multiple Sst ephys features was surprising, given we saw the poor overlap in associated genes between different ephys features of Sst (Fig. 2a), and poor overlap of ephys features between cell types (Fig. ...
Citations
... Single cell assays have been developed to capture diverse aspects of genome regulation, including gene expression 1,2 , chromatin accessibility 3 , and methylation profiling 4,5 , among others 6,7 . These single modality assays that capture a single data type have been widely deployed on a variety of tissues and species to catalog cell types and states [8][9][10][11][12][13][14][15][16] , identify genomic features that activate at specific steps along cellular trajectories [17][18][19][20][21][22][23] , and infer regulatory networks cataloging interactions of genes, open chromatin regions or methylation sites [24][25][26] . A common step of single cell data analysis is cell state inference: the inference of a low dimensional representation of a single cell data modality, that is subsequently used for 2D data visualization 27,28 , clustering to identify discrete cell types and states, and trajectory inference tasks 29,30 . ...
... Allaway et al., E13 mouse scRNA-seq, scATAC-seq, and scMultiome dataset processing Single cell paired and unpaired multiomic datasets from the E13 mouse medial ganglionic eminence (MGE) region were obtained 65 . After cell QC steps based on quantity metrics including the number of RNA counts (300-30,000), the number of RNA features (200-6000), the number of ATAC counts (300-60,000), the number of ATAC features (300-30,000), percentage of ribosomal genes (2%-30%), nucleosome signal (<1.5) and TSS enrichment (3)(4)(5)(6)(7)(8)(9)(10)(11)(12)(13)(14)(15), performed using Seurat (v4.3.0) and Signac (v1.10.0) in R, the dataset contained 11,258 cells in the unimodal scRNA-seq dataset, 24,589 cells in the unimodal scATACseq dataset, and 5,308 cells with paired multiomic profiles. For the analysis, we adopted the strategy used by the authors of the original paper, focusing exclusively on postmitotic cells. ...
... We processed the transcriptomic data using Seurat (v4.3.0) in R and excluded the Meis2 cell type, as it has been reported as a distant branch preceding the major glutamatergic and GABAergic split 100 . We also removed the aspiny dendrite type due to their substantial difference in transcriptomics profiles from the remaining GABAergic cell population 15 . For the ephys data, we performed feature extraction on raw recording data stored in NWB files using the AllenSDK (v0. ...
Multimodal single-cell assays profile multiple sets of features in the same cells and are widely used for identifying and mapping cell states between chromatin and mRNA and linking regulatory elements to target genes. However, the high dimensionality of input features and shallow sequencing depth compared to unimodal assays pose challenges in data analysis. Here we present scPair, a multimodal single-cell data framework that overcomes these challenges by employing an implicit feature selection approach. scPair uses dual encoder-decoder structures trained on paired data to align cell states across modalities and predict features from one modality to another. We demonstrate that scPair outperforms existing methods in accuracy and execution time, and facilitates downstream tasks such as trajectory inference. We further show scPair can augment smaller multimodal datasets with larger unimodal atlases to increase statistical power to identify groups of transcription factors active during different stages of neural differentiation.
