NIH Researchers Develop AI Tool to Predict Cancer Drug Responses at Single-Cell Level

In contrast to bulk sequencing, single-cell RNA sequencing provides high-resolution data at the individual cell level.

In a recent study, researchers at the National Institutes of Health (NIH) have unveiled an artificial intelligence (AI) tool capable of predicting cancer drug responses using data from individual cells within tumors.

Published in Nature Cancer on April 18, 2024, the study conducted by the National Cancer Institute (NCI) suggests that this innovative approach could revolutionize cancer treatment by enabling more precise matching of patients with effective drugs.

Unlocking Single-Cell Insights

Current methods of matching patients to drugs rely on bulk sequencing of tumor DNA and RNA, which averages the genetic data from all cells within a tumor sample.

However, tumors are composed of multiple cell types, including various subpopulations known as clones, which may respond differently to drugs. This variability can contribute to treatment resistance and suboptimal responses in patients.

In contrast to bulk sequencing, single-cell RNA sequencing provides high-resolution data at the individual cell level.

This technology offers insights into the genetic profiles of different cell clones within tumors, potentially leading to more effective drug targeting. However, the high cost and limited availability of single-cell gene expression data in clinical settings have posed challenges to widespread adoption.

AI-Powered Prediction Models

In their study, NIH researchers utilized a machine learning technique called transfer learning to develop AI models capable of predicting drug responses based on bulk RNA sequencing data, which is more readily available.

To enhance their accuracy, the models were then fine-tuned using single-cell RNA sequencing data.

The AI models accurately predicted drug responses for 44 FDA-approved cancer drugs, including single drugs and combinations.

By analyzing published data from patients with multiple myeloma and breast cancer, the researchers demonstrated that even if a single clone within a tumor was resistant to a drug, the patient would not respond to that drug.

Moreover, the AI model successfully forecasted the development of drug resistance in patients with non-small cell lung cancer.

Towards Clinical Implementation

While the study showcases the potential of AI-driven precision oncology, the researchers underscore the need for wider availability of single-cell RNA sequencing data to improve the accuracy of the technique.

To facilitate the adoption of this approach, the researchers have developed a research website and a guide for utilizing the AI model, named Personalized 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 was conducted at NCI's Center for Cancer Research under the supervision of Eytan Ruppin, M.D., Ph.D.

As the leading agency in the National Cancer Program, NCI spearheads efforts to reduce cancer prevalence and enhance patient care.

Through extensive research and training initiatives, NCI supports groundbreaking studies aimed at tackling cancer from all angles.

As the nation's premier medical research agency, NIH plays a pivotal role in funding and conducting research across various disciplines, including cancer biology and treatment.

The development of the AI tool by NIH researchers harnessing the power of AI and single-cell sequencing technology holds promise for improving patient outcomes and transforming cancer care in the future.