NIH researchers have unveiled a groundbreaking AI tool with the potential to revolutionize cancer treatment by offering more precise matching of drugs to patients.

In a proof-of-concept study published in Nature Cancer, scientists from the National Cancer Institute (NCI) leveraged single-cell RNA sequencing data to develop an AI model capable of predicting individual tumor responses to specific drugs.

Traditionally, cancer drug matching has relied on bulk sequencing, averaging the genetic makeup of all cells within a tumor sample. However, tumors consist of diverse cell populations, or clones, which may respond differently to treatment. Single-cell RNA sequencing provides higher resolution data, offering insights into individual clones' responses.

Using a machine learning technique called transfer learning, researchers trained the AI model on bulk RNA sequencing data before fine-tuning it with single-cell RNA sequencing data. Remarkably, the model accurately predicted responses to both single drugs and combinations across various cancers.

Testing the approach on patient data revealed crucial insights. Even if a single clone proved resistant to a drug, the patient wouldn't respond, underscoring the need for precise targeting. The AI model also forecasted resistance development in patients with non-small cell lung cancer, showcasing its predictive prowess.

While the technique's accuracy hinges on the availability of single-cell RNA sequencing data, researchers have developed a research website and guide for using the AI model, named Personalised Single-Cell Expression-based Planning for Treatments In Oncology (PERCEPTION), with new datasets.

Led by Alejandro Schaffer, Ph.D., and Sanju Sinha, Ph.D., the study underscores the potential of AI-driven precision medicine in oncology, marking a significant step towards tailored cancer therapies.