First-of-its-kind AI tool predicts patient response to cancer drugs at single-cell resolution

With over 200 types of cancer and each cancer being unique, continued efforts to develop precision oncology treatments remain challenging. Most of the focus has been on developing genetic sequencing assays and analyzes to identify mutations in cancer driver genes, and exploring potentially effective treatments against those mutations. .

However, many, if not most, cancer patients do not benefit from these early targeted therapies. According to a new study published April 18, 2024 in the journal natural cancerfirst author Sanju Sinha, Ph.D., assistant professor in the Sanford Burnham Prebys Cancer Molecular Therapy Program, and senior authors Eytan Ruppin, M.D., Ph.D., and Alejandro Schaffer, Ph.D., National Cancer Institute, part of the National Institutes of Health (NIH). and colleagues describe a first-of-its-kind computational pipeline to systematically predict patient responses to anticancer drugs at single-cell resolution.

This new artificial intelligence-based approach, called personalized single-cell expression-based treatment planning, or perception, in oncology leverages the utility of transcriptomics, the messengers expressed by transcription factors, genes that carry out transmission and transformation. We delve deeper into the usefulness of studying RNA molecules. Leverage DNA information.

“Tumors are complex, evolving beasts, and single-cell resolution allows us to address both of these challenges,” says Sinha. “With PERCEPTION, we can harness the rich information within single-cell omics to understand the clonal architecture of tumors and monitor the emergence of resistance.” (In biology, omics refers to the intracellular ).

Being able to monitor the emergence of resistance is the most exciting part for me. This may allow us to adapt to the evolution of cancer cells and even change treatment strategies. ”

Dr. Sanju Sinha, Assistant Professor, Sanford Burnham Prebys Cancer Molecular Therapy Program

Sinha and colleagues built PERCEPTION using transfer learning, a branch of AI.

“Limited single-cell data from clinics was our biggest challenge. AI models require large amounts of data to understand diseases. ChatGPT collects large amounts of text from the internet. It’s like you need data.”

PERCEPTION pre-trains the model using bulk gene expression from publicly available tumors. Single cell data from cell lines and patients were then used to calibrate the model, albeit to a limited extent.

PERCEPTION was successfully validated by predicting response to monotherapy and combination therapy in three recently published independent clinical trials in multiple myeloma, breast cancer, and lung cancer.

In both cases, PERCEPTION correctly stratified patients into responder and non-responder categories. In lung cancer, the development of drug resistance as the disease progresses has also been detected, a remarkable discovery with great potential.

Sinha says PERCEPTION is not yet ready for use in the clinic, but the approach shows that single-cell information can be used to guide treatment. He hopes to encourage adoption of this technology in clinics to generate more data and use that data to further develop and refine the technology for clinical use.

“The quality of predictions is only as good as the quality and quantity of the underlying data,” says Sinha. “Our goal is to create clinical tools that can systematically and data-drivenly predict the treatment response of individual cancer patients. We hope that these findings will continue to evolve as more data and such research I hope it will be promoted soon.”