Revising the ortholog conjecture in cross-species comparison of scRNA-seq data

The cross-species comparison of expression profiles uncovers functional similarities and differences between cell types and helps refining their evolutionary relationships. Current analysis strategies typically follow the ortholog conjecture, which posits that the expression of orthologous genes is most similar between species. However, the extent to which this holds true at different evolutionary distances is unknown. Here, we systematically explore the ortholog conjecture in comparative scRNA-seq data. We devise a robust analytical framework, GeneSpectra, to classify genes by expression specificity and distribution across cell types. Our analysis reveals that genes expressed ubiquitously across nearly all cell types exhibit strong conservation of this pattern across species, as do genes with high expression specificity. In contrast, genes within intermediate specificity fluctuate between classes. As expected, ortholog expression becomes more divergent with increased species distance. We also find an overall correlation between similarity in expression profiles and sequence conservation. Finally, our results allow identifying gene classes with highest probability of expression pattern conservation that are most useful for cell type alignment between species. Calibrating reliance on the ortholog conjecture for individual genes, we thus provide a comprehensive framework for the comparative analysis of single cell data.