From biological complexity to clinical precision: A first-in-class single-patient–level AI strategy for precision drug repurposing in relapsed, refractory, or metastatic cancers using COSMIC mutational signatures with blast-crisis CML as a model.

Abstract

e15117 Background: Introduction: Relapsed, refractory & metastatic cancers (RRM) are resistant to treatment due to extreme inter-patient heterogeneity & clinical limitations of treatment paradigms [1]. COSMIC mutational signatures capture underlying mutagenic pathways & provide a biologically integrated treatment framework that transcends single-gene alterations [2]. We hypothesized that integrating COSMIC signatures with artificial-intelligence (AI) algorithms could enable clinically actionable, single-patient precision drug repurposing in RRM [3]. Blast-crisis chronic myeloid leukemia (BC-CML) a genomically unstable & fatal disease was used as a proof-of-concept example [4]. Methods: Whole-exome sequencing was performed & COSMIC mutational signatures were extracted using SigProfilerExtractor to quantify dominant mutational processes at individual-patient resolution [1,5]. Unsupervised machine learning identified biologically coherent mutational signatures in each patient [3]. AI-guided drug prioritization was performed using PanDrugs, integrating mutations, signatures, and pharmacogenomic evidence to generate patient-specific, clinically actionable drug rankings [2]. Results: Each BC-CML patient exhibited a unique COSMIC signature profile, reflecting distinct biological drivers of leukemic progression and therapeutic vulnerability. Signature-defined biology mapped patients to hallmark pathways (Table). AI-based integration prioritized distinct repurposable FDA/EMA-approved therapies for each patient. No two patients shared an identical therapeutic profile, demonstrating true N-of-1 precision beyond cohort- or cluster-level classification. Conclusions: Discussion:This is a first-in-class, single-patient precision oncology framework shifting therapeutic decision-making from static mutations to dynamic mutational processes [1,6]. Using BC-CML as a model, this strategy is broadly applicable to RRM enabling rational drug repurposing when standard therapies fail supporting real-time, patient-level clinical decision-making [4,7]. References 1: Bobo M, et al. Int J Infect Dis. Jan 2026 2: Zhang L, et al. NPJ Precis Oncol. 2025 Dec 20 3: Li W et al. Discov Oncol 2026 Jan 12 4: Bhat M. Exp Cell Res. 2026 Feb 15 5: Ochi Y et al. Nat Commun. 2021 6: Wan Z. Adv Sci. 2025 7: Noor WD et al. Expert Rev Hematol 2026 Jan 11. CML Patient IDs COSMIC Signature(s) Core Pathway Top Repurposed Drugs 1–4 S3 / S5 HR deficiency, checkpoint failure PARP inhibitors (olaparib), CDK/MDM2 inhibitors 5–7 S1 Epigenetic dysregulation IDH inhibitors (enasidenib), PARP/ATR inhibitors 8–9 S2 Replication stress ATR inhibitors (AZD6738), antimetabolites 10–12 S13 / S18 Oxidative & inflammatory stress JAK inhibitors (ruxolitinib), NAC