In the realm of oncology, the quest for precision treatments tailored to individual patients has faced significant challenges. While genetic sequencing has been a focal point, the effectiveness of early targeted therapies remains limited for many cancer patients. A groundbreaking study, published in Nature Cancer, introduces PERCEPTION—a revolutionary AI-driven pipeline designed to predict patient responses to cancer drugs at the single-cell level.
Led by first author Sanju Sinha, Ph.D., and senior authors Eytan Ruppin, M.D., Ph.D., and Alejandro Schaffer, Ph.D., the research harnesses the power of transcriptomics to delve into the intricate landscape of cancer biology. By analyzing messenger RNA molecules expressed by genes at single-cell resolution, PERCEPTION offers a comprehensive understanding of tumor heterogeneity and resistance mechanisms.
"The emergence of resistance monitoring capability is particularly promising," notes Sinha, emphasizing PERCEPTION's potential to adapt treatment strategies to cancer evolution.
Developed using transfer learning—a subset of AI—PERCEPTION overcame challenges posed by limited clinical single-cell data. By leveraging bulk-gene expression data for pre-training and fine-tuning with available single-cell data, the model demonstrated remarkable predictive accuracy across multiple clinical trials for various cancer types.
From monotherapy to combination treatments, PERCEPTION accurately stratified patients into responder and non-responder categories, even capturing the evolution of drug resistance in lung cancer. While not yet primed for clinical implementation, Sinha envisions PERCEPTION as a pivotal step toward personalized cancer treatment.
As Sinha underscores, the predictive prowess of PERCEPTION hinges on the breadth and quality of data. The goal is to cultivate a robust clinical tool capable of systematically forecasting treatment responses in individual cancer patients—a transformative advancement poised to revolutionize oncology practice.
More: https://medicalxpress.com/news/2024-04-ai-tool-responses-cancer-therapy.html
