Neoepitope prediction integrated with DNA repair gene profiling: mutation list, peptide–HLA predictions, network links and composite-score components

Data and code underlying the manuscript “Neoepitope Prediction Integrated with DNA Repair Gene Profiling for Precision Oncology” (Computational Biology and Chemistry, CBAC-D-26-04047). This deposit contains the complete analysis underlying the paper: the mutation list, every peptide–HLA binding prediction for MHC class I and class II, the mutation–peptide links, the per-peptide network metrics, and the code that produces the composite score and the tables. Contents. 1,419 missense sample-level calls across 29 DNA-repair genes from TCGA BRCA, COAD and OV (01, 02); 53,683 peptide windows forming the mutation–peptide edge list (03); 429,464 MHC class I predictions across eight HLA alleles (04); 105,495 MHC class II predictions across five HLA-DRB1 alleles (05); per-peptide network degrees and hub labels for all 33,405 unique peptides (06); gene-level and multi-mutation summaries (07, 08); the ranked peptide list (09); and the analysis scripts in code/. SCHEMA.csv lists every file with its full column names, and README.md documents each file. Key definitions. Degree is the number of distinct parental sample-level calls generating a peptide; hubs are peptides of degree greater than five (836 peptides, 2.5% of 33,405). The Network_Strength and epitope_sharing columns are prediction-row counts, that is degree multiplied by eight alleles. The 1,419 calls correspond to 885 distinct genomic variants across 877 tumour samples, so all mutation counts are recurrence counts. Provenance. Somatic variant calls and expression data derive from TCGA and are redistributed in derived form consistent with TCGA data-use terms. Class I affinities were predicted with NetMHCpan 4.1 and MHCflurry 2.0 orchestrated through pVACseq; class II with the IEDB class II tools.