Previous studies often identify molecular markers of “transcriptomic types” that do not readily inform cell phenotypes. We instead focused on analyzing gene ensemble profiles (i.e. gene families) encoding proteins that constitute cellular modules (Hartwell et al., 1999) (i.e. macromolecular machines, signaling complexes), which more readily explain and predict cell phenotypes. Leveraging substantial phenotypic information of PCPs as an assay, MetaNeighbor allowed us to systematically screen all gene families and rank their ability to discriminate PCPs. This enabled us to discover a rather small set of functionally related gene families, which likely shape the I/O communication patterns of PCPs.
