What is HRD? Homologous Recombination Deficiency (HRD) refers to impaired DNA double-strand break repair through the homologous recombination pathway. In tumors with HRD, alternative and error-prone DNA repair pathways become relied upon, creating a vulnerability that can be exploited therapeutically. Mutations in BRCA1, BRCA2, and other homologous recombination genes are the most well-known causes of HRD.
PARP inhibitors and HRD: PARP inhibitors (such as olaparib, rucaparib, and niraparib) exploit the concept of synthetic lethality with HRD. When PARP inhibition blocks base excision repair in a cell that already cannot use homologous recombination, the cell accumulates lethal DNA damage. PARP inhibitors have demonstrated significant clinical efficacy in BRCA-mutated ovarian and breast cancers, and olaparib has received FDA approval for BRCA-mutated pancreatic cancer as maintenance therapy after platinum-based chemotherapy.
Need for broader HRD biomarker identification: Only a minority of pancreatic cancer patients carry germline BRCA1/2 mutations. However, HRD may occur through additional mechanisms including somatic mutations in BRCA1/2, mutations in other homologous recombination genes (PALB2, RAD51, ATM), and epigenetic silencing of repair genes. Identifying reliable biomarkers of the HRD state beyond BRCA mutation status could expand the population eligible for PARP inhibitor therapy and improve patient selection.
HRD scoring and dataset: The study used publicly available pancreatic adenocarcinoma (PAAD) datasets from TCGA and GEO, with HRD scores calculated using established genomic scar-based scoring methods that measure genome-wide loss of heterozygosity, large-scale state transitions, and telomeric allelic imbalance - all signatures of impaired homologous recombination. Patients were classified as HRD-high or HRD-low based on these scores, creating two groups for gene expression comparison.
Weighted Gene Co-expression Network Analysis (WGCNA): WGCNA was applied to identify gene modules - clusters of genes whose expression levels are highly correlated across samples. Modules significantly associated with the HRD status were identified by correlating module eigenvalues (representative expression levels) with HRD scores. Hub genes within the HRD-associated modules were identified as the most connected nodes in the co-expression network, representing potentially key regulators of the HRD state.
LASSO regression for feature selection: Least Absolute Shrinkage and Selection Operator (LASSO) regression was applied to the candidate genes identified from WGCNA to select the most informative subset for HRD prediction. LASSO penalizes large regression coefficients, effectively driving many coefficients to zero and retaining only the genes that provide unique predictive information about HRD status. This method reduces the risk of overfitting when many correlated gene features are considered simultaneously.
Random forest for validation and importance ranking: Random forest was applied as a complementary machine learning approach to rank genes by their importance for classifying HRD status. Genes identified by both LASSO and random forest as informative were considered the most robust biomarker candidates, as they survived two independent feature selection approaches with different mathematical formulations.
Three key biomarkers identified: The integrated WGCNA, LASSO, and random forest analysis converged on three genes as the strongest HRD-associated biomarkers in pancreatic cancer: CKS1B (CDC28 Protein Kinase Regulatory Subunit 1B), HJURP (Holliday Junction Recognition Protein), and TPX2 (Targeting Protein for Xklp2). All three genes were overexpressed in HRD-high tumors and showed consistent patterns across independent validation cohorts.
Biological roles of the identified genes: CKS1B is involved in cell cycle regulation and ubiquitin-mediated protein degradation, with known roles in cancer cell proliferation. HJURP is a chaperone protein for CENPA histone variant involved in centromere assembly and chromosomal stability - functions directly relevant to the genomic instability characteristic of HRD tumors. TPX2 is a spindle assembly factor involved in mitosis, and its overexpression correlates with chromosomal instability and aggressive tumor behavior across multiple cancer types.
Prognostic significance: Higher expression of all three genes was significantly associated with worse overall survival in pancreatic cancer patients, consistent with their biological roles in promoting cell cycle progression and chromosomal instability. Kaplan-Meier analysis and multivariate Cox regression supported these genes as independent prognostic markers beyond standard clinical variables such as tumor stage and resection status.
Correlation with immune infiltration: The study also found that expression of these HRD biomarker genes correlated with specific immune cell infiltration patterns in pancreatic tumors. HRD-high tumors showed altered natural killer cell and T cell infiltration compared to HRD-low tumors, suggesting that the HRD state may influence tumor immunogenicity and potentially sensitivity to immune checkpoint inhibitor therapy.
Linking biomarkers to drug sensitivity: To assess the clinical utility of the identified HRD biomarkers for PARP inhibitor treatment selection, the study analyzed drug sensitivity data from cancer cell line databases (GDSC - Genomics of Drug Sensitivity in Cancer). Cell lines with high expression of CKS1B, HJURP, and TPX2 showed significantly greater sensitivity to PARP inhibitors including olaparib and niraparib compared to cell lines with low expression of these genes.
Multi-gene signature performance: A composite biomarker signature combining the expression levels of all three genes (CKS1B, HJURP, TPX2) as a multi-gene score showed stronger prediction of PARP inhibitor sensitivity than any single gene alone. This is consistent with the general principle in cancer biomarker research that multi-gene panels provide more robust and biologically comprehensive predictions than individual markers.
Potential to expand PARP inhibitor eligibility: The analysis suggested that a subset of pancreatic cancer patients without BRCA mutations might be identifiable as HRD-high based on the three-gene expression signature and could potentially benefit from PARP inhibitor therapy. This is clinically significant because it would expand the theoretical patient population eligible for PARP inhibitor trials beyond the 5-10% of PDAC patients with germline BRCA mutations.
Expanding the HRD-targeted therapy population: The identification of CKS1B, HJURP, and TPX2 as expression-based HRD biomarkers offers a potential strategy to expand PARP inhibitor candidacy beyond genetically defined HRD (BRCA mutations). If validated prospectively, patients classified as HRD-high by the gene expression signature could be enrolled in clinical trials of PARP inhibitors as a precision medicine approach to a disease with very limited targeted therapy options.
Integration into clinical decision frameworks: Future clinical implementation would require development of a standardized assay - such as an RT-PCR or RNA in situ hybridization panel - that could measure CKS1B, HJURP, and TPX2 expression levels from standard FFPE tumor biopsy samples routinely collected at diagnosis. The computational analysis pipeline would then generate a risk score guiding PARP inhibitor consideration.
Study limitations: The study was entirely bioinformatics-based, using retrospective dataset analysis without prospective clinical trial validation. The cell line-based PARP inhibitor sensitivity data represents an early-stage preclinical evidence level. Clinical validation in pancreatic cancer patient cohorts treated with PARP inhibitors, correlating the three-gene signature with actual treatment outcomes, is the essential next step before these biomarkers can influence clinical practice.