Method summary. Figure 1A: Data preprocessing. High-confidence TF target genes of four primary classes are collected from several datasets (Steps 1-2) and merged into composite lists with four extra classes for multiple lines of evidence (Step 3). Process-specific gene lists (Step 4) and TF target genes are assembled into a regulatory matrix (Step 5) such that unrelated genes are assigned to an additional "baseline" class. Figure 1B: TF significance tests with multinomial logistic regression models. For a given TF, the alternative model H₁ is a univariate multinomial regression model that associates response (process genes) and one predictor (TF target genes), such that TF targets are linearly associated to probabilities of process gene classes (Step 6). The null model H₀ associates response (process genes) to their relative frequency in the dataset (Step 7). Log-likelihood ratio test measures if H₁ provides a better fit to data than the simpler H₀ model (Step 8). All TFs are subject to independent testing (Step 9) and subsequent multiple testing correction (Step 10). TF, transcription factor knockout strain; ChIP, chromatin immunoprecipitation; TF, transcription factor; TFBS, transcription factor binding site.

Overview of yeast TF dataset. Figure 2A: Distribution of TF target genes from gene expression data (red, green), TF binding data (yellow, orange), and combined evidence (blue). Figure 2B: Distribution of regulator classes in the TF dataset.TF, transcription factor knockout strain; TF, transcription factor; TFBS, transcription factor binding site.

Cell cycle TF prediction. Figure 3A: Significance scores for 46 predicted TFs for cell cycle genes (FDR p ≤ 0.05). Red bars represent nine core cell cycle TFs and blue bars show secondary cell cycle TFs. Figure 3B: Predicted phase specificity for cell cycle TFs. Cell cycle phases are shown clockwise from top right, and TF activities across phases are shown in concentric circles (not drawn to scale). Intensity of red tone shows predicted proportion of TF activity from scaled regression coefficients. TF names are coloured according to their known phase of action, confirming the agreement between known and predicted functions of TFs. Figure 3C: Contribution of regulatory evidence in recovering 13 known cell cycle TFs. Bar color denotes class of evidence and bar height shows proportion of weight assigned to this class. Figure 3D: Performance and robustness comparison of m:Explorer and other methods. X-axis corresponds to proportion of genes presented to method, and Y-axis shows performance with the Area Under Curve statistic for 18 cell cycle TFs (AUC, 95% confidence intervals for 100 runs). m:Explorer values are plotted in wider, black lines. Numbers at the right end of curves reflect method performance with the full yeast dataset of 6253 genes. AUC, area under curve; G1, gap-1; G2, gap-2; M, mitosis; S, synthesis; TFs, transcription factors; TFBS, transcription factor binding site.

Quiescence (G₀) regulation and prediction of candidate TFs. Figure 4A: G₀ in S. cerevisiae occurs when a saturated culture is depleted of nutrients. As exponential growth stops with the exhaustion of glucose, cells pass diauxic shift and switch to slower respiratory growth. Growth is halted in post-diauxic phase, cells differentiate into quiescent and non-quiescent populations and enter G₀. Culture viability starts decreasing rapidly as G₀ progresses. Two sets of process-specific genes were used in independent m:Explorer runs: genes expressed during diauxic shift (yellow box), and genes expressed in G₀ cultures (blue box, G₀ maintenance). Figure 4B: Venn diagram for cell cycle genes, diauxic shift genes, and G₀ maintenance genes. Statistically significant enrichment is observed between diauxic shift and quiescence genes (Fisher's exact test, p = 0.0005, 62 genes). Figure 4C: Log-scaled TF scores from diauxic shift and G₀ maintenance predictions (top - TFs 1-50, bottom - TFs 51-97). Bar color represents fraction of final score attributed to G₀ profile (diauxic shift - yellow; G₀ maintenance - blue). First 12 TFs highlighted with asterisks were selected for experimental testing. TF, transcription factor.

Experimental validation of G₀ TFs. Figure 5A: Viability of ΔTF strains in contrast to wildtype and negative controls, shown as a hierarchically clustered heat-map. Strains are shown vertically and time-points in days horizontally, coloured cells denote log fold change of viability (red - increased viability, blue - decreased viability). Asterisks denote statistical significance of fold change (linear regression, LR test, FDR p ≤ 0.05). Inviable strains with viability less than CFU = 0.005 are labeled with the symbol #. Time-point on day one corresponds to exponentially growing cells, measured seven hours after inoculation. Figure 5B: High-resolution viability curves of the first 72 hours of the time-course covering exponential growth, diauxic shift and post-diauxic phase. A subset of ΔTF strains (coloured solid lines), as well as wildtype strains (black solid lines) and controls (dashed and dotted lines) are shown. Figure 5C: G₀ viability curve of super-wildtype strain Δbas1. Figure 5D: Viability curve of G₀-essential strain Δ tup1. Figure 5E: Viability curve of positive control strain Δmip1. Figure 5F: Viability curve of negative control strain Δtup1. TF, transcription factor knockout strain; FDR, false discovery rate.

Gene Ontology and pathway analysis of G₀ TFs.Quiescence genes with differential expression in G₀ TF knockout microarrays were studied with ranked enrichment analysis in g:Profiler. Statistically significant non-redundant functional Gene Ontology categories are shown (hypergeometric test, FDR p ≤ 0.05). Black bars correspond to total number of G₀ genes in a given category, and coloured bars show the number of times these genes were up-regulated (red) or down-regulated (green) in related G₀ ΔTFs, according to knockout data. Enrichment p-value is shown in vertical colour strips on the left. Top: functional enrichments of quiescence genes associated to all G₀ TFs; middle: functional enrichments specific to TFs with reduced viability phenotype; bottom: Functional enrichments specific to WT+ TFs with increased viability phenotype. GO, Gene Ontology; TF, transcription factor.