Most cancer patients do not see the benefits of early targeted therapies. Now, scientists have introduced a new computational pipeline that can systematically predict how patients will respond to cancer drugs at a single-cell level.
With over 200 types of cancer, each unique in its own way, the task of developing precise oncology treatments remains challenging. The focus has mainly been on creating genetic sequencing tests or analyses to detect mutations in cancer driver genes. After that, attempts are made to find treatments that may be effective against those mutations.
nts do not benefit from these early targeted therapies. A new study, published on April 18, 2024, in the journal Nature Cancer, introduces a novel computational pipeline to predict patient response to cancer drugs at a single-cell resolution. The study’s first author is Sanju Sinha, Ph.D., an assistant professor in the Cancer Molecular Therapeutics Program at Sanford Burnham Prebys, with senior authors Eytan Ruppin, M.D., Ph.D., and Alejandro Schaffer, Ph.D., at the National Cancer Institute, part of the National Institutes of Health (NIH), and their colleagues. This new approach is called PERsonalized Single-Cell Expression-Based Planning.PERCEPTION, the innovative artificial intelligence-based method, delves further into the application of transcriptomics in oncology. Transcriptomics is the study of transcription factors, which are the messenger RNA molecules produced by genes. These molecules carry and convert DNA information into action. According to Sinha, “A tumor is a complex and evolving beast. Using single-cell resolution can allow us to tackle both of these challenges.” He adds, “PERCEPTION allows for the use of rich information within single-cell omics to understand the clonal architecture of the tumor and monitor the emergence of resistance.” In biology, omics refers to the sum of constituents within a cell.
Sinha expresses his enthusiasm for the potential to monitor and adapt to cancer cell evolution and modify treatment strategies through the use of PERCEPTION. This AI model, developed using transfer learning, overcame the challenge of limited single-cell data by pre-training its models with published bulk-gene expression from tumors. The importance of large amounts of data for the AI model to understand the disease is likened to the need for ChatGPT to have vast amounts of text data scraped from the internet.
Data from both cell lines and patients, although limited, was utilized to calibrate the models. PERCEPTION was effectively confirmed by accurately predicting the outcomes of single and combination therapies in three separate clinical trials for multiple myeloma, breast, and lung cancer. In each instance, PERCEPTION correctly categorized patients as either responders or non-responders. Additionally, it was able to identify the development of drug resistance in lung cancer as the disease advanced, which is a significant finding with promising implications. Sinha notes that while PERCEPTION is not yet suitable for clinical use, the approach has potential.ach indicates that individual cell data can guide treatment decisions and he wants to promote the use of this technology in clinics to generate more data. This data can then be used to improve and enhance the technology for clinical purposes.
Sinha states that the accuracy of the prediction depends on the quality and quantity of the data it is based on. The goal is to develop a clinical tool that can systematically and data-driven predict the response to treatment for cancer patients. It is hoped that these findings will lead to more data and further research in the near future.
Other authors involved in the study are Rahulsimham Vegesna, Sumit Mukherjee, Ashwin V. Kammula, Saugato Rahman Dhruba, Nishanth Ulhas Nair, Peng Jiang, Alejandro Schäffer, Kenneth D. Aldape, and Eytan Ruppin from the National Cancer Institute (NCI) collaborated with Wei Wu, Lucas Kerr, Collin M. Blakely, and Trever G. Biovona from the University of California, San Francisco; Mathew G. Jones and Nir Yosef from the University of California, Berkeley; Oleg Stroganov and Ivan Grishagin from Rancho BioSciences; Craig J. Thomas from the National Institutes of Health; and Cyril H. Benes from Harvard University.
This research was partially supported by the Intramural Research Program of the NIH, NCI, and NIH grants R01CA231300, R0.1CA204302, R01CA211052, R01CA169338 and U54CA224081.
Â
