Diffuse large B-cell lymphoma (DLBCL) is the most common non-Hodgkin lymphoma (NHL) and is a particularly aggressive disease with poor outcomes in the relapse setting. Patients are treated with chemotherapy and rituximab, an antibody targeting CD20 (R-CHOP). Approximately 50% of patients will experience disease progression and unfortunately, the majority of patients with relapsed/refractory disease (rrDLBCL) will die from their lymphoma. The standard of care for fit patients in the relapse setting is salvage chemotherapy as a bridge to autologous stem cell transplant (ASCT). More recently, chimeric antigen receptor (CAR) T cell therapy has provided improved survival to rrDLBCL over ASCT as a second line therapy. However, the majority of rrDLBCL will still progress following therapy and improving the outcomes for these patients is an unmet clinical need. We need to better understand the biological mechanisms responsible for relapsed disease and the ways by which malignant cells resist specific therapies. The benefit of therapies such as CART is improved when tumor burden is low at the time of infusion, and many patients are ineligible for treatment due to toxicity concerns. Therefore, there is a growing need to identify patients who are refractory to treatment as early as possible so there may be an earlier therapeutic intervention and possibly improved survival.Newer genetic classification of DLBCL, e.g. the LymphGen algorithm, has identified as many as seven genetic subtypes of DLBCL with potential therapeutic targets. However, additional mutations are acquired via selective pressure of therapy that may also contribute to rrDLBCL. While these mutations can be detected from peripheral blood via targeted sequencing of plasma circulating tumor DNA (ctDNA), the enrichment of certain mutations at relapse has not fully explained the chemo-resistant nature of DLBCL. We hypothesized that there may be functional defects in apoptotic activation in primary lymphoma that is not explained by genomic events. Using a technique called BH3 profiling, we aimed to expand on our understanding of apoptosis in DLBCL and compare it to other lymphomas. In addition, we also hypothesized that early treatment failure can be identified by serial profiling of ctDNA during therapy. Therefore, we applied a custom panel of 194 genes specific to rrDLBCL to 547 samples from 237 patients to determine if ctDNA can be used to identify patients that require early intervention and/or a change of therapy.BH3 profiling of 124 primary NHL tumor samples led to the identification of a subset of DLBCL that has remarkable reductions in apoptotic response and an apparent defect in pro-apoptotic proteins (BAX/BAK) needed for cell death. We also show that the majority of NHL depend on anti-apoptotic proteins MCL1 and BCL2 for survival, indicating a potential rationale for their targeting in certain NHL. Secondly, we used a novel method of disease monitoring involving the analysis of ctDNA to assess the mutational status of patients as they progress through therapy. We identified that ctDNA levels both pretreatment and mid-therapy are higher in patients who will be refractory to treatment. Moreover, mutations in TNFAIP3 and BTG2 were associated with improved survival, while rrDLBCL had a large number of mutations involved in cell survival and immune evasion (BCL2, MYC, B2M, CD83). Monitoring ctDNA in advance of relapse, we were able to detect ctDNA as early as 7 months before clinical presentation of relapsed disease, showing the use of ctDNA monitoring in detecting relapse even in patients with response to therapy.Taken together, we have identified both functional and genetic defects in DLBCL that contribute to disease progression. The resulting work supports to the potential utilization of apoptotic analysis and ctDNA monitoring in future patient care to aid in treatment management and subsequently, optimization of patient response