Abstract Diffuse large B-cell lymphoma (DLBCL) is a heterogeneous B-cell malignancy that consists of no fewer than 5 genetic subtypes. Cell lines have been a central model system for DLBCL but fail to fully recapitulate the diversity of genetic alterations that have been observed in large patient cohorts. Approximately 40% of DLBCLs will be refractory to or relapse following first-line R-chemo regimens, with many of these patients destined to receive CD19 CAR T-cells in the third line or later. Due to the relatively recent introduction of CD19 CAR T-cells into standard of care for relapsed/refractory DLBCL, there is a paucity of models in the CAR T relapsed setting. Together, these factors led us to embark upon an effort to build a substantive DLBCL patient-derived xenograft (PDX) repository that encompasses the major DLBCL genetic subtypes and includes CAR T relapsed models. From September 2019 to March 2022, we implanted 231 biopsies with diagnoses of DLBCL (n=120), HGBL (n=17), PMBL (n=2), other high grade B-cell lymphomas (n=87) and other low grade B-cell lymphoma (n=5). Our approaches were modeled from those utilized by the Public Repository of Xenografts (PRoXe: Townsend et al., Cancer Cell 2016), employing implantation of intact core needle biopsies into the renal capsule of NSG mice. Of 206 models that have been monitored for up to 6 months, we have generated PDX tumors from 52 tumors originating from 50 patients (25.2% take rate). The average time from implantation to signs of illness was 8 weeks (2 min, 22 max). The average number of preceding lines of therapy for these models was 3 (0 min, 8 max), with 22 models being generated from biopsies taken subsequent to progression from CD19 CAR T-cells. This included paired biopsies from two patients generated from biopsies taken immediately prior to receiving and following progression from CD19 CAR T-cells. All PDX models were confirmed to be EBV negative by EBER-ISH and subjected to histopathologic and immuno-phenotypical examination to compare their characteristics to an alternative core taken alongside the biopsy used to generate the PDX model. We additionally performed extensive molecular characterization, including RNA, whole-exome (WES), low-pass whole-genome (lpWGS), ATAC and ChIP sequencing for 7 histone post-translational modifications using the Mint-ChIP approach. Epigenetic characterization will be reported at a later date. RNA-seq confirmed both ABC (n=16) and GCB (n=24) subtypes. WES identified cardinal genetic features of all major genetic subtypes of DLBCL, including coexistent MYD88 L265P and CD79B mutations (MCD subtype); EZH2, CREBBP and KMT2D mutations (EZB subtype); NOTCH2 truncating mutations (N2 subtype); SGK1 and TET2 mutations (ST2 subtype); and NOTCH1 mutations (N1 subtype). In summary, we have established and characterized a large LBCL PDX model repository to serve as a resource for mechanistic and pre-clinical studies that will improve our understanding of lymphoma etiology and mechanisms of response/resistance to therapy. Citation Format: Haopeng Yang, Vida Ravanmehr, Jordan C Showell, Estela M Rojas-Neira, Jared M Henderson, Jason R Westin, Michael R Green. Building of a large B-cell lymphoma (LBCL) patient-derived xenograft (PDX) repository [abstract]. In: Proceedings of the Third AACR International Meeting: Advances in Malignant Lymphoma: Maximizing the Basic-Translational Interface for Clinical Application; 2022 Jun 23-26; Boston, MA. Philadelphia (PA): AACR; Blood Cancer Discov 2022;3(5_Suppl):Abstract nr A36.