Integrative Multiomic and Deep Learning Profiling of KRAS Mutant and Wildtype Genomes in Pancreatic Tumorigenesis

Pancreatic ductal adenocarcinoma exhibits high intertumoral heterogeneity, with KRAS mutations as the dominant oncogenic driver. However, a subset of these tumors retains a wild-type KRAS genotype yet progresses through alternative molecular mechanisms. Deciphering how these divergent tumors converge on shared malignant outcomes is crucial for precision oncology. We conducted an integrative multiomic analysis across whole-exome sequencing, RNA-Seq, methylation profiling, and proteomic data. Gene regulatory network (GRN) reconstruction, centrality analysis, T-Test and functional clustering were performed. A deep neural network model was developed for stratifying and validating KRAS-mutant and wild-type tumors based on identified transcriptomic signatures. KRAS-mutant tumors harbored canonical hotspot mutations (G12D, G12V, G12R). In contrast, KRAS-wildtype co-occurring with disruptive variants in TP53, CDKN2A, and SMAD4 highlighting a genomically unstable landscape and displayed enrichment of damaging variants in GNAS, with upregulation of alternative pathways involving hormonal and neuropeptidergic signaling. Multiomic integration identified TFAP2A (LFC: 5.124), and LCN2 (LFC: 4.835) as hyperactive effectors in KRAS-mutants and wildtype, supported by high mRNA and hypomethylated values. Wild-type tumors showed marked upregulation of CARTPT (LFC: 7.535) suggesting adaptive reliance on endocrine and immune modulation. Network analysis revealed seven core functional modules, with CAV1(LFC:2.25) emerging as central hubs in therapy resistance and EMT-metabolic signaling and found to have expression in both lung and liver metastasis. Sustained expression of CAV1 and the conserved nature of GRN seed node variants reinforce their contribution to metastatic evolution. A DNN trained on GRN-prioritized biomarkers achieved AUC = 0.94, accurately stratifying KRAS status and correlating with patient survival (HR = 0.46, p = 0.0021). Despite differing upstream mutations, KRAS-mutant and wild-type PDAC tumors converge on shared transcriptional and epigenetic programs that promote malignancy. These findings emphasize the role of regulatory convergence in tumor evolution and GRN-defined hubs as robust, mutation-agnostic therapeutic targets.