Grant Number:	1R01CA307720-01A1 Interpret this number
Primary Investigator:	Uno, Hajime
Organization:	Dana-Farber Cancer Inst
Project Title:	Advanced Integration and Validation of Predictive Models for Germline Pathogenic Variant Identification
Fiscal Year:	2026

Advancements in sequencing technologies have transformed genetic testing for cancer prevention and treatment, yet germline testing remains significantly underused. Despite guideline recommendations and its potential to guide therapy and identifying hereditary cancer risks, fewer than 10% of cancer patients undergo germline testing, with even lower uptake in some racial and ethnic groups. This underuse leaves many patients with undiagnosed pathogenic germline variants (PVs) and missed opportunities for targeted interventions. A key contributor is the reliance on predictive models developed primarily in non-Hispanic White cohorts, which may misestimate genetic risk when applied more broadly. To address these gaps, this project will develop and validate an integrative prediction model to identify cancer patients who would benefit from germline testing. Our approach builds on established tools, including the PREMMplus model for clinical risk prediction, the LOss of Heterozygosity and Genomic Instability Classification (LOHGIC) for tumor sequencing analysis, polygenic risk scores (PRS) for assessing inherited susceptibility, and genetic ancestry data. Using large datasets from multiple U.S. Cancer centers, we will create a robust, integrated prediction model for germline PV risk assessment that is highly generalizable. We propose the following Specific Aims: #1. Validate the PREMMplus model to ensure its predictive performance in identifying PVs across racial, ethnic, and sex-based subgroups using the large, potentially non-homogeneous data from over 40,000 patients from four sites (Vanderbilt, Dana- Farber Cancer Institute [DFCI], Memorial Sloan Kettering Cancer Center [MSKCC] and City of Hope [COH]; #2. Develop an integrated prediction model combining clinical, demographic, family history, tumor sequencing, and genetic ancestry data to enhance PV identification using the data from approximately 28,300 patients from DFCI, MSKCC and COH; and #3. Validate the integrated model using independent datasets (approximately 5,150 patients from DFCI, MSKCC and COH) to confirm its generalizability and effectiveness. The expected outcome is a scalable, accessible tool that improves the identification of cancer patients with germline PVs, supports broader access to germline testing, and strengthens precision in cancer treatment and prevention strategies. By integrating state-of-the-art predictive methodologies for germline PV assessment, this project addresses a critical public health gap in genetic testing and cancer care.





None