5057 Background: mCRPC with high neoantigen burden due to DNA mismatch repair deficiency (MMRd) and somatic hypermutation, or CDK12 mutations, may respond best to immune checkpoint inhibition (ICI). CCTG IND.232 suggested that combination ICI with DT has efficacy in a subset of mCRPC but should be biomarker-directed. In this sub-study (SS) of the PC-BETS master protocol, we explored DT in ARPI-resistant mCRPC stratified by circulating tumor DNA (ctDNA) analysis. Methods: Pts had ECOG PS 0–1, evaluable disease, biochemical or radiographic progression, prior ARPI ± cytotoxic chemotherapy (max 1 in castrate resistant setting). Genomic screening tested plasma ctDNA and matched leukocyte DNA via deep targeted sequencing with a prostate cancer-specific panel including coding regions and introns of selected mismatch repair genes and estimating tumor mutational burden. Only pts with evidence of ctDNA ≥1% were eligible. Pts with a positive biomarker (BM+) on ctDNA were assigned to a specific SS by a molecular tumor board (MTB); BM- pts were randomized between SS. SS- F tested DT in 2 cohorts of pts: cohort 1: BM+ pts had either somatic hypermutation (HM) ± concomitant MMR gene alterations, or CDK12 mutations, cohort 2: BM- pts without these alterations, in a 2-stage design. Primary endpoint was clinical benefit rate (CBR), defined by PSA50 response, RECIST CR/PR, or SD ≥12 weeks. Pts received T 225mg IV once on cycle 1, day 1 and D 1500mg IV day 1 every 4 weeks. Results: From January 2020 to February 2024, 25 pts were enrolled: 15 and 10 to cohort 1 and 2, respectively. Median age was 69y (63-84). Five pts had liver mets, 15 pts had had prior cytotoxics. 9 pts in cohort 1 had HM and the remainder had CDK12 mutations only. Median N cycles given was 4 (1-45). 12 pts had a delayed or interruption of DT dosing. The most common related AEs were fatigue (36%), rash (36%), and diarrhea (32%). CBR was seen in 53% of BM+ pts: all also had PSA response and significantly higher median ctDNA% (34% vs 5%; 5-69%); no patient selected as BM+ due to CDK12 mutations had CBR, while 8 of 9 pts selected based on HM had CBR. Conclusions: Liquid biopsy biomarker-informed treatment with DT demonstrated very promising efficacy in mCRPC pts with HM and merits further evaluation. CDK12 mutations were not predictive of CBR. Toxicities experienced were characteristic of ICI. Clinical trial information: NCT03385655 . BM + BM - CBR 8 (53%) 0 Median ctDNA% 25% 12.5% TTP-PSA* (mo; 95% CI) 6.7 (1-NR) 1.9 (1.1-NR) mOS** (mo; 95% CI) 15.2 (11.8-NR) 7.5 (2.1-NR) *PSA time to progression; **median overall survival.