Weinshilboum, R.

Abstract Background: The increasing volume and diversity of clinical genomic reports pose a significant challenge across healthcare institutions for the interpretation and proper implementation of precision oncology research. Reports from multiple different vendors (e.g., Invitae, Ambry Genetics, Foundation Medicine) are typically PDFs and exhibit substantial heterogeneity in panel design, gene coverage, and reporting standards, which hinders efficient retrospective data mining and patient cohort identification within the electronic medical records. We sought to develop a framework to interrogate all of these reports. Methods: We extracted genomic reports from patients with breast cancer treated with neoadjuvant chemotherapy and developed MolHarmonizer, a novel, scalable framework leveraging Python and Gemini LLMs, designed to process and harmonize genomic data from disparate multi-vendor reports. Gemini LLMs are employed explicitly for robust information extraction, normalization, and structuring of key genomic features, transforming unstructured data into a unified, queryable dataset. Results: Our MolHarmonizer framework successfully processed 1,147 genomic reports from 1703 breast cancer patients (2006-2023) from 23 different companies, demonstrating robust capability to extract and standardize critical actionable biomarkers. Data sources included Invitae (n=554), Ambry Genetics (n=189), Natera (n=95), Mayo Clinic (n=88), Tempus (n=63), Guardant Health (n=47), and Foundation Medicine (n=37), with others contributing less than 20 reports. Of the samples, 827/1147 (72.1%) were germline (blood/saliva). A majority of the patients were tested using the panels due to a personal/family history (n=925). Overall, 413/1147 (36.0%) reports identified at least one mutation. For breast cancer, 75 reports showed BRCA1/2 mutations (37 BRCA1, 37 BRCA2, and one patient with both BRCA1 and 2). Other mutations identified included: PIK3CA (n=40), TP53 (n=96), PTEN (n=21), ESR1 (n=13) and AKT1/2 (n=8). Conclusion: MolHarmonizer, a powerful framework leveraging Gemini LLMs, effectively addresses genomic data heterogeneity by automating biomarker extraction and harmonization. This enables rapid cohort identification and deep retrospective analyses for clinical insights, biomarker discovery, understanding disease history, facilitating novel pattern discovery, e.g., predicting BRCA1 mutations from WSI, and accelerating research within our neoadjuvant BC cohort. Future plans include expanding to include over 20,000 breast cancer patients, developing a user-friendly chatbot, and ensuring inter-institutional adaptability for a variety of complex diseases. Citation Format: Krishna Rani Kalari, Xiaojia Tang, Thanmayee Boyapati, Tanya L. Hoskin, Sumathilatha Myla, Sumedha G. Penheiter, Richard M. Weinshilboum, Liewei Wang, Hamid R. Tizhoosh, Karthik Vikram Giridhar, Matthew P. Goetz, Judy C. Boughey. Development of an LLM framework for analysis of heterogeneous breast cancer patients genomic reports [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2026; Part 1 (Regular Abstracts); 2026 Apr 17-22; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2026;86(7 Suppl):Abstract nr 2756.

Abstract Background: Tumor-infiltrating lymphocytes (TILs) in HER2+ and triple-negative breast cancer (BC) are prognostic and associated with therapy response; however, there are few data regarding distribution and significance of TIL values in ER+ HER2-negative (HER2-) MBC patients (pts) treated with CDK4/6 inhibitors (CDK4/6i). CDK4/6 inhibition may promote antitumor immunity by production of type III interferons and tumor antigen presentation, and by inhibition of regulatory T cells (Goel, Nature 2017). Further, the addition of immune checkpoint therapy to palbociclib numerically prolonged PFS in the 2nd line (2L) setting (Mayer, JCO 2024). We aimed to investigate the prognostic role of TILs derived from fresh metastatic tumor biopsies in pts with ER+HER2- MBC prior to starting CDK4/6i and endocrine therapy (ET) in the first line (1L) and 2L setting. Methods: PROMISE (NCT0281902) is a multicenter, prospective study that enrolled women commencing the CDK4/6i palbociclib (Pb) with letrozole (1L) or fulvestrant (2L) for HR+HER2- MBC. The primary goal was to identify novel genomic variants and pathways associated with PFS. Here, we performed a secondary analysis to quantitate stromal TILs using protocol mandated pretreatment tumor biopsies and determine their association with PFS. Using the Kaplan-Meier method, we reported median PFS (mPFS) and 3-year PFS percentage (%) (95% confidence intervals [CIs]) by TILs [≥20% vs <20%] separately within 1L and 2L. As sample TIL values were generally concentrated at zero (27/58 = 46.6%), coupled with wide spread of sample TIL values ≥20% [range: 20%- 80%], we applied a ≥20% threshold for PFS analysis (high vs low TIL group) as described by Luen, Salgado et al. Lancet Oncol 2017. Statistical significance was based on the Wilcoxon test of the null hypothesis that there was no difference in the PFS curves by TILs. PFS is reported at median follow-up of 21 months. Results: Pretreatment tumor biopsies from 58/68 (85.3 %) pts were evaluable (42 in 1L and 16 in 2L). Median [Q1, Q2] TIL values were higher in 1L (5% [0, 10]) vs 2L (0% [0, 15]). The most common biopsy sites were bone (44.8%), lymph node (17.2%), liver (17.2%) and breast (12%). Overall, TILs frequency (>20%) by the 4 most common biopsy sites were bone (2/20, 10%), lymph node (4/7, 57%), liver (2/6, 33%), and breast (1/7, 14%). In 1L, although PFS was not significantly different (P = 0.237), mPFS (44 vs 34 months) and 3-year PFS 75.0%, [95% CI: 50.3%, 100%] vs 46.4% [95% CI: 31.0%, 69.5%]) were longer in pts with high versus low TILS, respectively. In 2L, the 3-year PFS was numerically similar comparing high and low TILs groups, (25.0%, [95% CI: 4.6%, 100%] vs 22.5% [95% CI: 6.7%, 76.1%]) with no significant difference in PFS (P = 0.761). Conclusions: In pts with ER+HER2- MBC treated with Pb and ET, PFS was numerically longer comparing high vs low TIL values in 1L but not 2L settings. Future studies evaluating TILs in larger datasets are needed to validate these findings. Genomic analysis (DNA/RNA) and its association with TIL levels is underway and will be presented at the meeting. Funding: The Center for Individualized Medicine, Mayo Clinic; Mayo Clinic Breast Specialized Program of Research Excellence (SPORE P50CA116201); Mayo Clinic Cancer Center (CA15083-40A2); TEMPUS (https://www.tempus.com/) and Conquer Cancer Foundation of ASCO (Career Development Award 2018, Dr. O’Sullivan) Citation Format: C. C. O'Sullivan, A. M. Moyer, A. Nassar, D. M. Zahrieh, V. Suman, X. Tang, K. J. Thompson, R. C. Hentz, J. J. Harrington, A. Moreno-Aspitia, D. W. Northfelt, T. C. Haddad, S. Chumsri, P. P. Peethambaram, K. J. Ruddy, K. V. Giridhar, R. A. Leon-Ferre, R. Weinshilboum, L. Wang, K. R. Kalari, M. P. Goetz. Tumor-infiltrating lymphocytes in advanced ER-positive, HER2-negative breast cancer prior to treatment with palbociclib and endocrine therapy: a retrospective analysis of the PROMISE study [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 PS1-10-22.