In a significant advancement for computational biology, researchers at MIT have unveiled a new machine-learning framework called PottsMPNN. This innovative system aims to improve the success rate of protein design by shifting focus away from merely reproducing sequences found in nature.
The function of a protein is inherently linked to its structure, which is determined by the sequence of amino acids that compose it. Traditional methods for designing novel proteins typically follow a two-step process: first, establishing the desired structure, and then employing a machine-learning framework to generate a variety of sequences that could potentially adopt that structure. However, many amino acid sequences can fold into the same structure, complicating the design process.
New Perspectives on Protein Design
“For years, the field has measured success by asking whether a model can reproduce the protein sequence that evolution happened to select — our work shows that this isn’t the best metric for protein design,” states Amy E. Keating, head of the Department of Biology and senior author of a recent paper published in PNAS.
PottsMPNN incorporates physical principles that govern protein structure and stability, enhancing its ability to generate sequences and predict how mutations will impact protein stability. This model offers a more nuanced understanding of the sequence-energy landscape, which describes the relationship between amino acid identity and protein stability.
Breaking Away from Tradition
By integrating PottsMPNN into the protein design pipeline, researchers can create structurally feasible proteins with sequences that do not resemble any existing native proteins. As graduate student and lead author Foster Birnbaum explains, “If we’re thinking about a completely novel, designed structure, there would be no native sequence to compare it to.” The focus is on how likely the generated sequences are to fold into the desired structures and how well the model can predict the effects of mutations.
Birnbaum notes that the introduction of “noise” during training—variations added to protein structures—helps the model avoid mimicking native sequences too closely, thereby increasing the diversity of possible structures. Furthermore, PottsMPNN captures interactions between amino acids through a pairwise distribution, enhancing its modeling accuracy.
Future Implications
As the field of protein design evolves, the implications of PottsMPNN are profound. “Once we can design any protein we want, that enables us to do a potentially scary amount of biological engineering,” Birnbaum remarks. The framework not only paves the way for designing new-to-nature proteins for various applications but also lays a stronger foundation for future advancements in the field.
Keating concludes, “Our methods move the field toward designing useful new-to-nature proteins for diverse applications while providing a stronger foundation for future advances.”
This article was produced by NeonPulse.today using human and AI-assisted editorial processes, based on publicly available information. Content may be edited for clarity and style.
