Healthcare (Commonwealth Union) – The application of artificial intelligence (AI) continues to play a significant role across all areas of research. As AI continues to advance the ability to analyze large amounts of data often exceeding multiple decades, within a short period of time. This has shed light on many unseen areas of research; however, many developers have warned that verification AI analyzed data is essential.
Scientists at Baylor College of Medicine have created a new AI model that uncovers how changes in proteins connect genetic mutations to various diseases. The tool, named DeepMVP and detailed in Nature Methods, surpasses earlier models and could open doors to developing new treatments.
Dr. Bing Zhang, senior author of the study, professor at the Lester and Sue Smith Breast Center, and member of Baylor’s Dan L Duncan Comprehensive Cancer Center indicated that proteins carry out nearly every task in the body—building tissues, regulating metabolism, and defending against illness. These functions are often controlled by chemical changes that occur after proteins are produced, a process known as post-translational modification (PTM).
Such modifications include attaching chemical groups like phosphates or sugars, which determine how a protein functions, where it travels inside a cell, and how long it remains active. When PTMs malfunction, proteins may fail to work properly, contributing to conditions such as cancer, cardiovascular disease, or neurological disorders.
Mapping where PTMs occur helps scientists predict how mutations in these regions might alter protein activity and affect human health. For example, mutations in DNA can erase an existing PTM site, generate a new one, or disrupt nearby regions—ultimately reshaping how the protein operates.
“We developed DeepMVP, a computational model to predict where in a protein PTMs happen and which mutations in those locations can affect PTMs,” explained co-first author Dr. Chenwei Wang, a postdoc in the Zhang lab. “To train the model to recognize patterns in protein sequences that indicate PTM sites, we created the PTMAtlas, a curated compendium of known 397,524 PTM sites generated through systematic reprocessing of 241 public datasets. We focused on six common PTMs.”
PTMAtlas contains close to 400,000 PTM sites spread across thousands of human proteins. Unlike other databases, PTMAtlas is broader and more precise – it can forecast PTM sites in every human protein and even within viral proteins, including those from SARS-CoV-2. This demonstrates that DeepMVP is a highly effective tool for examining protein modifications.
DeepMVP surpassed eight comparable methods. When evaluated on its ability to predict how mutations influence PTMs using a curated dataset of 235 documented mutation-PTM pairs from published research, DeepMVP correctly identified the PTM site in 81% of instances and accurately determined whether the modification increased or decreased in 97% of cases.
Zhang indicated that they expect DeepMVP to be applicable to cancer, neurological disorders, and cardiovascular disease, helping speed up discoveries in genetics, oncology, and drug development.
Other contributors to the project include co-first authors Bo Wen and Kai Li, along with Ping Han, Matthew V. Holt, Sara R. Savage, Jonathan T. Lei, Yongchao Dou, Zhiao Shi, and Yi Li — all from Baylor College of Medicine.
This research received backing from the National Cancer Institute (NCI) CPTAC award U24CA271076, the Cancer Prevention and Research Institute of Texas award RR160027, support from the McNair Medical Institute at the Robert and Janice McNair Foundation, and NCI grant R01 CA271588. Additional assistance came from NVIDIA Corp., which provided a Titan Xp GPU for the study.
