Abstract 4000: A large-scale, multi-target deep learning model for virtual genomic profiling in colorectal cancer

Abstract Molecular profiling of tumor biopsies is the cornerstone of precision oncology, guiding therapy selection and prognostic assessment. Computational analysis of routine H&E slides offers a complementary approach that can accelerate insights and optimize downstream molecular testing. Recent advances in computational pathology have demonstrated that deep learning models can infer molecular features directly from digital H&E images, termed virtual genomics. However, most studies in colorectal cancer (CRC) have been constrained by limited cohort sizes and single-target prediction frameworks, restricting their clinical relevance and generalizability. Here, we present a transformer-based deep learning framework designed for large-scale, single- and multi-task virtual genomic profiling in CRC. Trained on 45,155 patients with matched digital H&E images, whole-exome sequencing (WES), and circulating tumor DNA sequencing for molecular recurrence monitoring, we validated all results on an external cohort from The Cancer Genome Atlas (TCGA, n=422 CRC patients). We embedded all images using H-optimus-0 and trained a transformer-based multiple instance learning aggregation head for downstream virtual genomic predictions. We trained individual models to predict genes found in the MSK-IMPACT505 cancer gene panel and a single model that predicts all genes simultaneously. Individual models predicted 379 mutated genes with an internal area under the receiver operating curve (AUROC)>0.7. We validated 254 of these genes within TCGA with an AUROC>0.7. Training a multi-task learning model to predict all genes simultaneously improved the AUROC for 85% of the genes. Based on NCCN guidelines in CRC, we further trained individual models to predict MSI status, BRAF V600E, KRAS G12D/V/G13D, and POLE/POLD1 exonuclease mutations with AUROCs of 0.96, 0.92, 0.83, and 0.83, respectively. For the BRAF, KRAS, and POLE/POLD1 mutations, we observed continued benefit from increasing cohort size from as small as 10 mutated cases up to the maximum positivity within our cohort, suggesting that we have not reached the full performance potential for these mutations. Finally, virtually-imputed BRAF mutations stratified minimal residual disease (MRD) risk better than the WES-derived genomic mutation status revealing that the model was capturing MAPK pathway activation rather than single mutation status. The virtual genomic algorithm advances virtual genomic profiling in CRC, demonstrating scalable and potentially generalizable prediction of hundreds of genomic alterations from routine histology. Beyond serving as a low-cost pre-screening tool to guide molecular testing, the virtually imputed biomarkers capture pathway-level activity that better correlates with clinical outcomes such as MRD risk, underscoring its potential to transform precision oncology through image-based genomic insight. Citation Format: Erik N. Bergstrom, Tinghui Wu, Michail Chatzianastasis, Thinh Tran, Aaron Rosenfeld, Robert Burns, Adham Jurdi, Helio Costa, Frank Zhang, . A large-scale, multi-target deep learning model for virtual genomic profiling in colorectal cancer [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2026; Part 1 (Regular Abstracts); 2026 Apr 17-22; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2026;86(7 Suppl):Abstract nr 4000.