Fig. 4: Key pathway activations and genes frequently implicated in DILI.

A Pathways are ranked by their ability to distinguish DILI vs. non-DILI compounds, measured by AUC. To quantify the strength of association between pathway dysregulation and predicted DILI risk, we show the Spearman correlation between pathway activation scores and DILI probability across compounds. This highlights transcriptional and metabolic stress pathways—such as nuclear receptor signaling, one-carbon metabolism, and bile acid regulation—as key contributors to DILI risk. B Compounds are ordered by their DILI risk probability, showing a correlation between DILI risk and the activation of pathways associated with liver injury. Pathway enrichment was performed using Enrichr (gseapy), based on a Fisher’s exact test with Benjamini–Hochberg correction. The most affected pathways include those critical to drug metabolism, oxidative stress and lipid homeostasis. Key pathways include Cytochrome P450 Oxidation, NRF2-ARE regulation for oxidative stress response, and Fatty Acid Beta-Oxidation, highlighting the interplay between detoxification, mitochondrial function, and energy metabolism. Additionally, nuclear receptor pathways such as Pregnane X Receptor (PXR) and Sterol Regulatory Element-Binding Protein (SREBP) signaling reveal disruptions in lipid and bile acid regulation. C Differential gene signatures were computed with DESeq2 using the Wald test, with Benjamini–Hochberg correction. Values are shown as –log10 adjusted p-values multiplied by the sign of the log-fold change. Top 100 genes most frequently upregulated in DILI cases include those related to drug metabolism (e.g., CYP1A2, CYP51A1, UGT1A8, AKR1C1, AKR1C2, AKR1C3), drug transport (e.g., SLC3A2), stress response (e.g., TXNRD1, SRXN1, GLRX, GLRX3, GCLM, HSP90AA1, HSP90AB1, HSPE1), and lipid metabolism (e.g., PLIN2, INSIG1, SREBF1, SCD, LPIN2, FADS2). Less commonly studied but significant genes include those involved in inflammation (e.g., GDF15, TNFRSF12A, S100A10, LITAF), autophagosome formation (e.g., MAP1LC3B, SQSTM1), and mitochondrial function (e.g., VDAC2, RAN, CHCHD10). D Top 100 genes most frequently downregulated in DILI cases include those involved in drug metabolism and detoxification (e.g., CYP4A11, CYP2E1, AKR1C4, GSTA1, GSTA2, UGT2B10, UGT2B15), drug transport (e.g., SLC10A1, SLC22A1, SLC22A7, SLC38A3, SLC27A5), and protein processing (e.g., ALB, AHSG, TTR, APOA1, APOE). Genes associated with amino acid metabolism and mitochondrial function (e.g., MAT1A, ARG1, CPS1, HMGCS2, ACAA1, ACAA2, ECHS1) also show significant downregulation. Key oxidative stress regulators and redox enzyme (e.g., CAT, ABAT, GNMT, and DHTKD1) are also reduced. Pathways related to lipid metabolism (e.g., CIDEB, ACSM2A, ACSM2B, ACSM5) and coagulation or inflammatory responses (e.g., SERPINF2, SERPINC1, SERPINA6, SERPINA10, FGB, FGG, FGA) are prominently affected. Notable genes with less established links to DILI but showing significant downregulation include ITIH4, PGLYRP2, BHMT, and GUCA2B, which could represent novel mechanisms or pathways contributing to the progression of liver injury. The changes in gene expression reflect early cellular alterations that may lead to DILI and highlight the complexity of DILI mechanisms, which encompass multiple aspects of liver function.