Skip to main content
. Author manuscript; available in PMC: 2022 Mar 29.
Published in final edited form as: Nat Comput Sci. 2021 Feb 22;1(2):120–127. doi: 10.1038/s43588-021-00030-1

Fig. 2 ∣. Cross-modal reconstructions capture cell-type-specific gene expression patterns and electrophysiological features.

Fig. 2 ∣

a, Gene expression levels averaged over samples reference taxonomy cell types, normalized per gene by the maximum value of each column. b, The cell-type specificity of different genes is captured well by cross-modal prediction of gene expression profiles from electrophysiological features (Pearson’s r = 0.89 ± 0.10, mean ± s.d. over cell types). c, A subset of electrophysiological features pooled by cell types shows analogous cell-type specificity. d, Cross-modal reconstructions of the electrophysiology features from gene expression profiles match the measured electrophysiology features (Pearson’s r = 0.98 ± 0.02, mean ± s.d. over cell types). Cell-type and feature-wise fidelity of reconstructions are quantified with Pearson’s r for each row and column in b and d as compared to ground truth in a and c.