Abstract Background: Germline microRNA-disrupting variants (mirSNPs) predict late radiation toxicity, which generally presents as fibrosis, in several cancer types. mirSNP signatures of toxicity are radiation fractionation (size of delivered dose) dependent, meaning that identifying patients at risk of toxicity from one form of radiation treatment affords the opportunity to choose safer alternatives for their management. As radiation is an integral part of breast cancer management, with several fractionation approaches gaining acceptance, we tested the hypothesis that mirSNP signatures could predict late radiation toxicity in these patients. Methods: DNA was isolated from buccal swabs from 121 breast cancer patients treated with conventionally fractionated radiation (CFRT, 38%) or moderately hypofractionated radiation (16-20 fractions, 62%) and tested for over 120 mirSNPs. Acute, chronic and late radiation toxicity as well as clinical factors were collected, including post-mastectomy radiation or implant, chemotherapy treatment, alcohol and smoking history, autoimmunity, and DNA repair mutant status. Most of the cohort was White and non-Hispanic (81.8%), with 11.5% of women White and Hispanic, and 6.6% of women non-White and non-Hispanic. We applied a previously developed mirSNP signature of late grade 2 CFRT radiation toxicity defined in prostate cancer patients 1. We then explored the addition of clinical factors to this signature to see if any improved predictive accuracy. Finally, we incorporated germline mirSNPs marginally associated with late toxicity (Fisher or Jonckheere-Terpstra p 0.1) in our breast cohort alongside the CFRT prediction in elastic net and boosted tree models. Results: The late CFRT radiation toxicity signature achieved robust predictive performance for late grade 2 toxicity in our breast patient cohort, with a specificity of 0.957, a positive predictive value (PPV) of 0.79, and an overall AUC of 0.734. Notably, all patients with grade 0 late toxicity were accurately identified as low risk, and the model predicted grade 3 late toxicity more accurately than grade 2 (60% vs. 30%, p=0.317 for difference in proportions). As expected, the late CFRT radiation toxicity signature showed no predictive power for predicting acute or chronic toxicity in our cohort, adding to our recent research demonstrating different biological pathways involved in radiation induced acute, chronic, and late toxicity 2. We found that none of the clinical factors improved the performance of the late CFRT radiation toxicity signature in our cohort, and in fact, the LOOCV AUC slightly decreased to 0.710. In contrast, adding just 6 additional mirSNPs, including MAT1A and KRAS, boosted the AUC to 0.810-0.837 and increased the sensitivity from 0.52 (CFRT alone) to 0.79 (CFRT + 6 SNPs). Interestingly, only 2 of these mirSNPs overlapped with the original CFRT model, and both MAT1A and KRAS variants appeared protective: the rate of late toxicity dropped from 38% to 4% (wild type to mutant MAT1A) and from 27% to 5% (wild type to heterozygous KRAS). Conclusions: These findings further highlight the power and generalizability of germline mirSNP signatures to meaningfully predict late radiation-induced toxicity, including in breast cancer patients. These are important steps towards meaningful biological insights into late radiation toxicity, as well as indicate the possibility of truly personalizing radiation recommendations in breast cancer.1.Kishan, A.U., et al., Germline variants disrupting microRNAs predict long-term genitourinary toxicity after prostate cancer radiation. Radiother Oncol, 2022. 167: p. 226-232.2.Kishan, A.U., et al., Validation and Derivation of miRNA-Based Germline Signatures Predicting Radiation Toxicity in Prostate Cancer. Clin Cancer Res, 2025. 31(12): p. 2530-2538. Citation Format: J. Weidhaas, K. McGreevy, S. McCloskey, J. Hegde, M. Steinberg, J. Baker, M. Dinome, A. Kusske, J. Glaspy, R. Callahan, K. McCann, M. Teshome, N. Mcandrew, A. Bardia, D. Telesca. Germline microRNA-based variants predicting late radiation toxicity in breast cancer abstract. In: Proceedings of the San Antonio Breast Cancer Symposium 2025; 2025 Dec 9-12; San Antonio, TX. Philadelphia (PA): AACR; Clin Cancer Res 2026;32(4 Suppl):Abstract nr PS1-07-09.
Paik et al. (Tue,) studied this question.