A Precision Oncology Platform to Support Genotype-Tailored Therapy Selection.

There is a critical need for a streamlined process to identify genotype-matched individuals eligible for enrollment into clinical trials and/or targeted therapies, as current methodologies face challenges in integrating diverse molecular data sources. We have developed a precision oncology platform to assist molecular tumor boards and community oncologists in reviewing patients' phenotypes, evaluating related knowledge, and identifying genotype-matched therapies.

Using Retrieval-Augmented Generation to Capture Molecularly-Driven Treatment Relationships for Precision Oncology.

Modern generative artificial intelligence techniques like retrieval-augmented generation (RAG) may be applied in support of precision oncology treatment discussions. Experts routinely review published literature for evidence and recommendations of treatments in a labor-intensive process. A RAG pipeline may help reduce this effort by providing chunks of text from these publications to an off-the-shelf large language model (LLM), allowing it to answer related questions without any fine-tuning. This potential application is demonstrated by retrieving treatment relationships from a trusted data source (OncoKB) and reproducing over 80% of them by asking simple questions to an untrained Llama 2 model with access to relevant abstracts.

Druggable genomic landscapes of high-grade gliomas.

Background: Despite the putatively targetable genomic landscape of high-grade gliomas, the long-term survival benefit of genomically-tailored targeted therapies remains discouraging. Methods: Using glioblastoma (GBM) as a representative example of high-grade gliomas, we evaluated the clonal architecture and distribution of hotspot mutations in 388 GBMs from the Cancer Genome Atlas (TCGA). Mutations were matched with 54 targeted therapies, followed by a comprehensive evaluation of drug biochemical properties in reference to the drug's clinical efficacy in high-grade gliomas. We then assessed clinical outcomes of a cohort of patients with high-grade gliomas with targetable mutations reviewed at the Johns Hopkins Molecular Tumor Board (JH MTB; n = 50). Results: Among 1,156 sequence alterations evaluated, 28.6% represented hotspots. While the frequency of hotspot mutations in GBM was comparable to cancer types with actionable hotspot alterations, GBMs harbored a higher fraction of subclonal mutations that affected hotspots (7.0%), compared to breast cancer (4.9%), lung cancer (4.4%), and melanoma (1.4%). In investigating the biochemical features of targeted therapies paired with recurring alterations, we identified a trend toward higher lipid solubility and lower IC50 in GBM cell lines among drugs with clinical efficacy. The drugs' half-life, molecular weight, surface area and binding to efflux transporters were not associated with clinical efficacy. Among the JH MTB cohort of patients with IDH1 wild-type high-grade gliomas who received targeted therapies, trametinib monotherapy or in combination with dabrafenib conferred radiographic partial response in 75% of patients harboring BRAF or NF1 actionable mutations. Cabozantinib conferred radiographic partial response in two patients harboring a MET and a PDGFRA/KDR amplification. Patients with IDH1 wild-type gliomas that harbored actionable alterations who received genotype-matched targeted therapy had longer progression-free (PFS) and overall survival (OS; 7.37 and 14.72 respectively) than patients whose actionable alterations were not targeted (2.83 and 4.2 months respectively). Conclusion: While multiple host, tumor and drug-related features may limit the delivery and efficacy of targeted therapies for patients with high-grade gliomas, genotype-matched targeted therapies confer favorable clinical outcomes. Further studies are needed to generate more data on the impact of biochemical features of targeted therapies on their clinical efficacy for high-grade gliomas.