Fig. 4

Robustness of regulon expression associated with transcription factor transcriptional adaptation. A Analysis schematic: is there differential regulon gene expression robustness after mutation of upstream transcription factors, associated with whether the transcription factors demonstrate transcriptional adaptation by paralogs? B Change in gene expression for each regulon gene after reference CRISPR transcription factor target mutation, compared against respective controls (log2(fold change) from DESeq2). Each point represents one downstream regulon member gene. Gray: regulon genes downstream of CRISPR target-paralog pairs not appearing to be involved in transcriptional adaptation. Red: regulon genes downstream of CRISPR target-paralog pairs with apparent transcriptional adaptation. Regulon genes called differentially expressed if DESeq2 adjusted p-value < 0.05 and abs(log2(fold change)) > 0.5. Fisher’s exact test for difference of odds of differential expression between the two groups of CRISPR target regulons, p = 1.82 × 10⁻⁴. C Gene expression distributions of a representative transcription factor that demonstrates possible transcriptional adaptation: SMAD4, in non-template controls (“control”, gray) and in SMAD4-guide-treated cells (“targeted”, orange). Fisher’s exact test p-values for difference in odds of cells being positive in control vs. targeted populations. Abundance values log(TPM + 1) reported by [52]. D Gene expression distributions of a representative top-100 paralog gene of a transcription factor that demonstrates possible transcriptional adaptation: SMAD1, a paralog of SMAD4, in non-template controls (gray) and in SMAD4-guide-treated cells (orange). Fisher’s exact test p-values for difference in odds of cells being positive in control vs. targeted populations. Abundance values log(TPM + 1) reported by [52]. E Gene expression distributions of three representative regulon genes of a transcription factor that demonstrates possible transcriptional adaptation: ID3, TNFRSF11B, and ID2, regulon genes of both SMAD4 and SMAD1, in non-template controls (gray) and in SMAD4-guide-treated cells (orange). Fisher’s exact test p-values for difference in odds of cells being positive in control vs. targeted populations. Bimodality coefficient calculated for non-zero subpopulation of each distribution. Abundance values log(TPM + 1) reported by [52]. F Change in gene expression for each regulon gene after reference CRISPR transcription factor target mutation, compared against non-template controls (i.e., percent-positive in CRISPR targeted cells minus percent-positive in non-template control-treated cells, for genes 75% or less in non-template controls). Each point represents one downstream regulon member gene. Gray: regulon genes downstream of CRISPR target-paralog pairs not appearing to be involved in transcriptional adaptation. Orange: regulon genes downstream of CRISPR target-paralog pairs with apparent transcriptional adaptation. Asymptotic test of difference in coefficient of variation for groups of unequal size, p = 0.0017. G Empirical distribution of standard deviation of change in percent-positive for downsampled (n = 49, same as transcriptional-adaptation group size) no-transcriptional-adaptation group regulon members without replacement, 1000 downsamples, in gray. Observed standard deviation of change in percent-positive transcriptional-adaptation group standard deviation to this distribution, in orange