Researchers at Baylor College of Medicine have developed an artificial intelligence model that sheds light on how protein modifications can connect genetic mutations to disease. The tool, called DeepMVP, was published in Nature Methods and is reported to outperform previous models used for similar purposes.
“Proteins are responsible for all the functions of the body, from growing tissues to regulating metabolism or fighting disease. Their functions are often regulated by modifications that take place after proteins are made through a process called post-translational modification (PTM),” said Dr. Bing Zhang, professor at the Lester and Sue Smith Breast Center and of molecular and human genetics at Baylor. He is also a McNair scholar and a member of Baylor’s Dan L Duncan Comprehensive Cancer Center.
Protein modifications, known as PTMs, involve chemical changes such as the addition of phosphates or sugars. These changes affect protein behavior, location within cells, and stability. Disruptions in PTMs have been linked to diseases including cancer, heart conditions, and neurological disorders.
Identifying where PTMs occur helps researchers predict how mutations might alter protein function in ways that impact health. DNA mutations can disrupt existing PTM sites or create new ones, leading to altered protein activity.
“We developed DeepMVP, a computational model to predict where in a protein PTMs happen and which mutations in those locations can affect PTMs,” said Dr. Chenwei Wang, co-first author and postdoctoral researcher 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.”
The newly created database—PTMAtlas—contains almost 400,000 identified modification sites across thousands of human proteins. This resource covers more proteins than previous databases and extends its predictions even to viral proteins like those from SARS-CoV-2.
DeepMVP was tested against eight existing tools for predicting mutation effects on PTMs using data from scientific literature covering 235 known mutation-PTM pairs. The results showed DeepMVP correctly predicted the site affected by mutation 81% of the time and accurately determined whether the effect increased or decreased 97% of the time.
“We anticipate that DeepMVP can be applied to cancer, neurological conditions and cardiovascular diseases and accelerate discoveries in genetics, cancer biology and drug development,” Zhang said. “The tool is freely available to researchers worldwide at https://deepmvp.ptmax.org/.”
Other contributors include Bo Wen and Kai Li (co-first authors), Ping Han, Matthew V. Holt, Sara R. Savage, Jonathan T. Lei, Yongchao Dou, Zhiao Shi and Yi Li—all affiliated with Baylor College of Medicine.
Funding for this study came from multiple sources: National Cancer Institute CPTAC award U24CA271076; Cancer Prevention and Research Institutes of Texas award RR160027; McNair Medical Institute at Robert and Janice McNair Foundation; NCI grant R01 CA271588; as well as hardware support from NVIDIA Corp., which provided a Titan XpGPU used during research.
