The introduction of poly (ADP-ribose) polymerase inhibitors (PARPi) into the management of ovarian cancer has transformed the treatment landscape for patients affected by this malignancy. However, as the use of PARPi expands into both frontline maintenance and recurrence settings, the emergence of drug resistance has become a significant clinical challenge in the treatment of these patients. Although platinum-based chemotherapy (PBC) and PARPi act through different mechanisms—PBC causes DNA damage while PARPi blocks its repair—both depend on the integrity of DNA damage repair (DDR) pathways, leading to overlapping mechanisms of resistance. Here, we review the key resistance mechanisms shared by PARPi and PBC, and then we discuss their clinical implications in the management of patients with ovarian cancer. We also examine clinical rationale supporting the hypothesis that prior PARPi exposure may reduce the efficacy of subsequent PBC in patients experiencing a disease recurrence. Furthermore, we review preliminary clinical data assessing the potential role of PARPi retreatment in patients who have previously progressed on PARPis.

Background: Mutations in the DNA polymerase delta 1 (POLD1) exonuclease domain cause DNA proofreading defects, hypermutation, hereditary colorectal and endometrial cancer, and are predictive of immunotherapy response. Exonuclease activity is carried out by two magnesium cations, bound to four highly conserved, negatively charged amino acids (AA) consisting of aspartic acid at amino acid position 316 (p.D316), glutamic acid at position 318 (p.E318), p.D402, and p.D515 (termed DEDD motif). Germline polymorphisms resulting in charge-discordant AA substitutions in the DEDD motif are classified as variants of uncertain significance (VUSs) by laboratories and thus would be considered clinically inactionable. We hypothesize this mutation class is clinically pathogenic. Methods: A review of clinical presentation was performed in our index patient with a POLD1(p.D402N) heterozygous proband with endometrial cancer. Implications of this mutation class were evaluated by a Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA)-guided systematic review, in silico analysis with orthogonal biochemical confirmation, and whole-exome and RNA sequencing analysis of the patient’s tumor and engineered cell lines. Results: Our systematic review favored a Mendelian disease mutation class associated with endometrial and colorectal cancers. In silico analysis predicted defective protein function, confirmed by biochemical assay demonstrating loss of nuclease activity. A POLD1-specific mutational signature was found in both the patient’s tumor and POLD1(p.D402N) overexpressing cell. Furthermore, paired whole-exome/transcriptome analysis of endometrial tumor demonstrated hypermutation and T cell-inflamed gene expression profile (GEP), which are joint predictive biomarkers for pembrolizumab. Our patient showed a deep, durable response to immune checkpoint inhibitor (ICI). Conclusion: Charge-discordant AA substitution in the DEDD motif of POLD1 is detrimental to DNA proofreading and should be reclassified as likely pathogenic and possibly predictive of ICI sensitivity.

Multimodal learning significantly improves cancer survival prediction using multiple data types for each patient, including clinical data, pathological images, and genomic data. However, existing works often assume that each data type contributes equally, neglecting potential conflicting or unreliable information between different modalities by simply concatenating features to form a fused representation. In addition, current frameworks are trained only to rank relative risk among patients and cannot quantify the actual likelihood of survival at specific time points, limiting their utility in real-world clinical settings. In this work, we propose the Modality-Aware Discrete-Time Survival (MADSurv) framework to address these gaps. First, it implements an uncertainty-aware attention mechanism in which modality-specific expert encoders learn both predictive features and their own confidence for intelligent data fusion. This approach leads to more robust and personalized predictions by dynamically focusing on the modalities that are most reliable for each individual patient. Second, in addition to a single overall risk score, MADSurv produces a sequence of survival probabilities for discrete yearly intervals. We evaluated our proposed method on five different cancer datasets. In addition to ranking patients’ overall risk using the concordance index, we also assessed the accuracy of the model’s survival probability estimates on each yearly milestone using the Brier score. Our experimental results demonstrate that MADSurv achieves superior and competitive performance compared to state-of-the-art methods.