A study published in Nature Biotechnology presents a promising new use for artificial intelligence: the design of small drug-like molecules capable of binding and destroying harmful proteins in the body, even without knowing their three-dimensional structure.
Technology can pave the way for treatments against diseases that are resistant to traditional therapies, such as certain cancers, neurological disorders and viral infections.
How the tool acts
- The tool, called PepMLM, was developed by researchers from McMaster, Duke and Cornell universities.
- Based on an algorithm originally created to understand human language, she was trained to interpret the “language” of proteins and design peptide drugs using only the amino acid sequence, without relying on structural models – an advance to reach proteins previously considered “untouchable”.
- “This approach changes the game,” says Pranam Chatterjee, work leader at Duke and today at the University of Pennsylvania.
- In tests, PepMLM generated peptides capable of binding and, in some cases, degrading proteins associated with cancer, Huntington’s disease and viral pathogens.
Avanço had a collective effort from research centers
The experiments involved multidisciplinary teams: at McMaster, Christina Peng demonstrated efficacy against toxic proteins of Huntington’s disease; at Cornell, Matthew DeLisa and Hector Aguilar worked with viral proteins; and at Duke, the model was created and validated.
The group is already developing new versions of AI, such as PepTune and MOG-DFM, to improve the stability and delivery of these peptides in the body. “We want a programmable therapeutic platform that starts from a sequence and results in a real medicine,” summarizes Chatterjee.
( fontes: olhar digital)
