Alternative splicing signatures to predict response from hyperthermic intra-peritoneal chemotherapy (HIPEC) treatment in ovarian cancer patients.

e17576 Background: HIPEC is associated with improved survival in epithelial ovarian cancer (EOC) patients, but no predictive biomarkers exist to guide optimal patient selection. Alternative splicing events (ASEs) have demonstrated prognostic value in multiple cancers, including EOC. We set out to identify response-associated ASEs as potential predictive biomarkers associated with HIPEC response in EOC patients. Methods: Ninety-one EOC patients who underwent HIPEC (2014-2022) with available pre-operative tumors at COH and CHU were identified. RNA was isolated from formalin-fixed paraffin-embedded samples, and whole-transcriptome libraries constructed. HIPEC response was defined by the following progression-free survival (PFS) cut-off values to distinguish good vs poor responders: 18 months in primary EOC patients (based on KGOG, CARCINO-HIPEC trials), and 12 months in recurrent EOC patients (based on MSK, CHIPOR HIPEC trials). We performed splice analysis using SplAdder and Bisbee to detect and quantify alternative splicing events (ASEs) from short read RNA-seq data from tumors. Significantly differentially ASEs between poor vs good responders were determined using Log-Likelihood Ratio (LLR), in Bisbee. MHC class I binding prediction of potential neopeptides generated by ASEs were computed using NetMHCpan. Results: A total of 56 EOC tumor samples were analyzed, after exclusion of 35 samples due to quality control or missing data. 57.1% were primary EOC, 42.9% recurrent EOC. Median follow up was 37.6 mo.; median PFS was 29.3 mo. in primary EOC and 26.0 mo. in recurrent patients. Good responders (n=39) had lower PCI, lower recurrence, and longer OS (95%CI: 47.9, NR), compared to poor responders (n=17), (95%CI: 24.1, NR). For the entire cohort, we identified 228 ASEs, of which 89 were predicted to produce a novel non-canonical protein sequence predominantly due to intron retention events (~80%). Thirty-six protein coding ASEs were significantly increased in good responders (LLR>20), with significantly positive enrichment in transcripts from Dead-Box family RNA Helicases (DDX5, DDX41, DDX27 and DDX3X), known to participate in RNA processing. By further selecting down ASEs with LLR>20 and Percent Spliced In (PSI) threshold of >5%, we retained 11 ASEs predicted to produce strong binders against MHC class I, including events in Serine/Arginine-Rich Splicing Factor 5 (SRSF5), a key protein involved in mRNA splicing. Conclusions: Alternative splicing analysis of RNA-Seq data from treatment naïve EOC tumors identified candidate protein coding ASE signatures involving the DDX family of RNA helicases to be associated with good response to HIPEC. Future plans include the development of a robust multi-panel ASE classifier expansion in a larger independent patient cohort to establish predictive treatment stratifications in HIPEC.