Machine learning (ML) is a key technology to enable accurate prediction of antibody-antigen binding, a prerequisite for in silico vaccine and antibody design. Two orthogonal problems hinder the current application of ML to antibody-specificity prediction and the benchmarking thereof: (i) The lack of a unified formalized mapping of immunological antibody specificity prediction problems into ML notation and (ii) the unavailability of large-scale training datasets. Here, we developed the Absolut! software suite that allows the parameter-based unconstrained generation of synthetic lattice-based 3D-antibody-antigen binding structures with ground-truth access to conformational paratope, epitope, and affinity. We show that Absolut!-generated datasets recapitulate critical biological sequence and structural features that render antibody-antigen binding prediction challenging. To demonstrate the immediate, high-throughput, and large-scale applicability of Absolut!, we have created an online database of 1 billion antibody-antigen structures, the extension of which is only constrained by moderate computational resources. We translated immunological antibody specificity prediction problems into ML tasks and used our database to investigate paratope-epitope binding prediction accuracy as a function of structural information encoding, dataset size, and ML method, which is unfeasible with existing experimental data. Furthermore, we found that in silico investigated conditions, predicted to increase antibody specificity prediction accuracy, align with and extend conclusions drawn from experimental antibody-antigen structural data. In summary, the Absolut! framework enables the development and benchmarking of ML strategies for biotherapeutics discovery and design.

Machine learning (ML) is a key technology for accurate prediction of antibody-antigen binding. Two orthogonal problems hinder the application of ML to antibody-specificity prediction and the benchmarking thereof: The lack of a unified ML formalization of immunological antibody specificity prediction problems and the unavailability of large-scale synthetic benchmarking datasets of real-world relevance. Here, we developed the Absolut! software suite that enables parameter-based unconstrained generation of synthetic lattice-based 3D-antibody-antigen binding structures with ground-truth access to conformational paratope, epitope, and affinity. We formalized common immunological antibody specificity prediction problems as ML tasks and confirmed that for both sequence and structure-based tasks, accuracy-based rankings of ML methods trained on experimental data hold for ML methods trained on Absolut!-generated data. The Absolut! framework thus enables real-world relevant development and benchmarking of ML strategies for biotherapeutics design.

Purpose: Homologous recombination (HR) deficiency (HRD) is one of the key determinants of PARP inhibitor response in ovarian cancer, and its accurate detection in tumor biopsies is expected to improve the efficacy of this therapy. Because HRD induces a wide array of genomic aberrations, mutational signatures may serve as a companion diagnostic to identify PARP inhibitor–responsive cases. Experimental Design: From the The Cancer Genome Atlas (TCGA) whole-exome sequencing (WES) data, we extracted different types of mutational signature–based HRD measures, such as the HRD score, genome-wide LOH, and HRDetect trained on ovarian and breast cancer–specific sequencing data. We compared their performance to identify BRCA1/2-deficient cases in the TCGA ovarian cancer cohort and predict survival benefit in platinum-treated, BRCA1/2 wild-type ovarian cancer. Results: We found that the HRD score, which is based on large chromosomal alterations alone, performed similarly well to an ovarian cancer–specific HRDetect, which incorporates mutations on a finer scale as well (AUC = 0.823 vs. AUC = 0.837). In an independent cohort these two methods were equally accurate predicting long-term survival after platinum treatment (AUC = 0.787 vs. AUC = 0.823). We also found that HRDetect trained on ovarian cancer was more accurate than HRDetect trained on breast cancer data (AUC = 0.837 vs. AUC = 0.795; P = 0.0072). Conclusions: When WES data are available, methods that quantify only large chromosomal alterations such as the HRD score and HRDetect that captures a wider array of HRD-induced genomic aberrations are equally efficient identifying HRD ovarian cancer cases.

Purpose: Poly (ADP ribose)-polymerase (PARP) inhibitors are approved for use in breast, ovarian, prostate, and pancreatic cancers, which are the solid tumor types that most frequently have alterations in key homologous recombination (HR) genes, such as BRCA1/2. However, the frequency of HR deficiency (HRD) in other solid tumor types, including bladder cancer, is less well characterized. Experimental Design: Specific DNA aberration profiles (mutational signatures) are induced by HRD, and the presence of these “genomic scars” can be used to assess the presence or absence of HRD in a given tumor biopsy even in the absence of an observed alteration of an HR gene. Using whole-exome and whole-genome data, we measured various HRD-associated mutational signatures in bladder cancer. Results: We found that a subset of bladder tumors have evidence of HRD. In addition to a small number of tumors with biallelic BRCA1/2 events, approximately 10% of bladder tumors had significant evidence of HRD-associated mutational signatures. Increased levels of HRD signatures were associated with promoter methylation of RBBP8, which encodes CtIP, a key protein involved in HR. Conclusions: A subset of bladder tumors have genomic features suggestive of HRD and therefore may be more likely to benefit from therapies such as platinum agents and PARP inhibitors that target tumor HRD.

Background: African American (AA) men have significantly higher mortality rates from prostate cancer (PC) than individuals of European ancestry (EA). Therapeutically targetable molecular differences may hold the potential to reduce this disparity. Objective: To investigate chromodomain helicase DNA-binding protein 1 (CHD1) deletion both as a cause of aggressive disease and therapeutic vulnerability in the prostate cancer of AA men. Design, setting, and participants: 91 AA and 109 EA prostate cancer cases were analyzed by fluorescence in situ hybridization (FISH) for the deletion of CHD1. Whole exome and whole genome sequencing data from prostate adenocarcinoma cases were analyzed for mutational signatures from AA and EA individuals. Outcome measurements and statistical analysis: Associations with biochemical recurrence were evaluated using Cox proportional hazard regression models. Association between mutational signatures and CHD1 deletion were assessed by Wilcoxon ranked sum tests. Results and limitations: Subclonal deletion of CHD1 is nearly three times as frequent in prostate tumors of men than in EA men. CHD1 deletion is associated with some of the homologous recombination deficiency associated mutational signatures in prostate cancer. In a cell line model CHD1 deletion induced 1-10 kb deletions resembling those induced by BRCA2 deficiency. CHD1 deficient cells showed markedly increased sensitivity to both talazoparib and the radiomimetic bleomycin. Conclusions: CHD1 is more frequently deleted in the prostate cancer of AA men. This deletion is both associated with and induces mutational signatures characteristic of BRCA2 deficiency. CHD1 deficient prostate cancer is more sensitive to talazoparib or bleomycin treatment. Patient summary: Subclonal deletion of CHD1 is more frequent in the prostate cancer of AA men and this could be one of the reasons behind more aggressive disease. CHD1 deletion, however, also constitutes a therapeutic vulnerability to the PARP inhibitor talazoparib. This treatment may significantly improve the outcome of disease in AA men.