<i>BRCA1/2</i> and <i>PALB2</i> short variants (SVs) contributed by clonal hematopoiesis (CH) in liquid biopsies (LBx) from patients with advanced pancreatic cancer (PC).

4183 Background: CH results from fitness-enhancing mutations in hematopoietic stem cells. Many CH somatic variants (SVs) are in cancer-associated genes, including ATM and CHEK2 , which do not have a homologous recombination deficiency (HRD) phenotype. SVs in well-established HRD-associated genes like BRCA1/2 and PALB2 also appear in white blood cells as CH, albeit more rarely. Herein, we report the prevalence of SVs of CH origin in these clinically relevant genes that confound results of PC liquid biopsies (LBx) and study their association with HRD in tissue biopsies from the same patients. Methods: This study uses a novel variant origin prediction algorithm to classify the origin of each SV detected by FoundationOneLiquid CDx as germline, tumor-somatic, or CH, using a combination of sample-specific and dataset-learned features including fragmentomics (trained and validated against white blood cell standard). 5,625 PC LBx sequenced during routine clinical care was used for broad prevalence data. A subset with matched tissue biopsies (TBx, n = 536) was used to compare SV detection and true HRD via HRDsig in TBx, a signature validated to predict response to PARP inhibitors across multiple cancer types. Results: Among 303 PC LBx with a BRCA1/BRCA2/PALB2 SV, 52 (17.2%) were predicted to be of CH origin. This percentage is larger than for other BRCA -associated canonical cancer types: prostate (14.1%, 139/980), breast (9.7%, 87/902), ovarian (7.0%, 14/200), but less than some non-canonical cancer types (Table). In PC patients with both LBx and TBx available, 29/536 (5.4%) had an SV in BRCA1/BRCA2/PALB2 detected in LBx: 8/29 in LBx only; 21/29 in both LBx and TBx. 19/29 were predicted germline by the algorithm, were all also detected in TBx, and 15 (79%) of these TBx were HRDsig+. 5/29 (17%) were predicted to be tumor-somatic; two of these were detected in TBx and one (20%) was HRDsig+. Predicted CH SVs were detected in another 5 LBx (4 BRCA2 ; 1 PALB2 ). None of these were detected in TBx and none of the tumors were HRDsig+. Conclusions: While the majority (58%) of BRCA1/BRCA2/PALB2 SV+ PC LBx harbored a predicted germline SV, 25% harbored tumor-somatic SV and 17% had SV exclusively predicted as CH-derived. Determining the cellular origin of BRCA1/BRCA2/PALB2 in PC is essential given the potential impact on treatment selection. Cancer type LBx, N n with SV in BRCA1/2/PALB2 (%) % Germline SV present % Tumor somatic SV present % No germline/tumor somatic SV, only CH SV present Lung 21456 928 (4.3) 25.5 51.3 23.1 Cholangiocarcinoma 1787 90 (5) 38.9 38.9 22.2 Pancreas 5625 303 (5.4) 57.8 25.1 17.2 Esophagus 1199 56 (4.7) 33.9 50.0 14.3 Prostate 13858 980 (7.1) 39.3 46.5 14.2 Colorectal 6639 391 (5.9) 15.3 73.4 11.0 Breast 11397 902 (7.9) 55.2 35.1 9.6 Ovarian 1464 200 (13.7) 66.0 27.0 7.0 Endometrial 795 84 (10.6) 20.2 73.8 6.0