NIH researchers develop AI tool for personalized oncology

On April 18, 2024, researchers from the National Institutes of Health (NIH) published a proof-of-concept study in Nature Cancer detailing a new artificial intelligence (AI) tool to match patients with the appropriate cancer drug based on their tumor RNA. The tool is called the Personalized Single-Cell Expression-Based for Treatments in Oncology (PERCEPTION).

Based on the data from this early-stage study, researchers theorize that single-cell RNA sequencing may be critical for effective and accurate cancer treatments in the future.

According to the NIH press release, current approaches in oncology rely on sequencing treatments based on generalized data from bulk sequencing of tumor DNA and RNA. However, these insights average all the tumor cells rather than looking individually at the types of cells in a tumor. Certain cell subpopulations may respond to drugs at different rates or in varying manners, which, in turn, means that patients will respond to sequences in different ways.

As a replacement for bulk sequencing, single-cell RNA sequencing offers higher-resolution data at the single-cell level, which can provide more detailed information on tumor composition and drug responses.

Armed with this data, oncologists may be able to identify the appropriate treatment or drug from the very first time. Oncology research has shown that the efficacy of the first-line treatment is one of the most critical factors contributing to cancer survivorship. With AI-based tools to optimize personalized oncology, cancer survivor rates may also improve.

In this study, the researchers built the AI model to match the sequencing data to 44 FDA-approved oncology drugs.

To test the theory, researchers analyzed data from 41 patients with multiple myeloma and 33 patients with breast cancer using varying drug combinations. According to the NIH press release, if one of the tumor cell subpopulations were resistant to a drug, the patient would not respond to the drug overall, regardless if the other cell subpopulations were not resistant.

While the data supports the use of AI models that analyze single-cell RNA sequencing to choose oncology drugs, more research is needed to assess these results further.