Unlocking Myeloma Insights With Generative AI

Praneeth Reddy Sudalagunta, PhD, research instructor in the department of cancer physiology at H. Lee Moffitt Cancer Center, discusses a novel generative artificial intelligence (AI) pipeline developed to help with the understanding and treatment of multiple myeloma.

For over a decade, the researchers, including Sudalagunta, worked to train the pipeline on a unique cohort of over 2000 patients with myeloma, which encompassed molecular profiles, functional drug sensitivity data from ex vivo models, and comprehensive clinical information. This wealth of data, currently housed in a Snowflake data warehouse, formed the foundation of the AI-driven approach.

“This large database, that we have is ingested into a Snowflake data warehouse currently, is combined with all the clinical lab data of the patients. All of that is being utilized using large language models. So essentially, what the large language model would do is that it would answer context-dependent questions that are prompted by the physician,” explains Sudalagunta in an interview with Targeted Oncology_^TM.

At the 2025 American Association for Cancer Research (AACR) Annual Meeting, Sudalagunta et al presented a poster on this sophisticated AI architecture. The AI integrates molecular, clinical, and functional data to create explainable predictive models and therapeutic strategies for multiple myeloma. In addition, the pipeline utilizes the Mistral 7B LLM, enhanced with Low-Rank Adaptation and Retrieval-Augmented Generation (RAG).

As part of the molecular data, investigators collect RNA sequencing and whole exome sequencing data, profiling the patient tumors molecularly. Looking at the functional data, experts co-cultured the patient-derived tumor cells in an ex vivo reconstruction of the bone marrow microenvironment. They went on to treat those tumor cells with various drugs, and measured the viability, representing patient-specific sensitivity to various drugs.

“What this does is actually provide deeply, biologically insightful information to physicians that can actually help them in clinical decisions,” he adds.