Identifying maximally informative signal-aware representations of single-cell data using the information bottleneck (opens in new tab)

Dubnov et al. develop an scRNA-seq-tailored framework based on the information bottleneck principle to extract compressed, optimally informative cellular representations with respect to cell labels (such as healthy vs. diseased). bioIB identifies biologically interpretable gene signatures that expose signal-specific biological processes and cellular phenotypes.