Categorization of identified CFD programs. (A) trackSeq detects functionally cooperating genes. STRING database queried for known protein–protein interactions of trackSeq candidates. Top 500 LysoHigh genes, top 500 LysoLow genes, and 1000 sets of 500 random genes as null distribution queried for interactions. (B) LysoHigh genes contain significantly more than random interactions. P values from null distribution. (C) trackSeq candidates form networks of 65 LysoLow (white) and 152 LysoHigh (green) genes (of top 500 candidates, respectively). (D) Functional enrichments in LysoHigh and LysoLow sisters. Expression bias: ratio of LysoHigh to LysoLow candidates in process, normalized by number of process genes in network. If >0, more LysoHigh candidates involved; if <0, more LysoLow candidates involved. (E) Functional predictions from network: HSC daughters modulate cell differentiation, protein catabolism, and signaling. LysoHigh network components suggest faster CC transition than in LysoLow sisters. Mean and variance of gene expression from Figure 2B, shown throughout figure. FDR, false discovery rate of process enrichment within network.