Arieh Warshel, a USC Distinguished Professor of Chemistry and a Nobel laureate, has spent his career trying to understand how enzymes respond to mutation and how those changes influence biological processes. His latest work focuses on a statistical framework known as maximum entropy, which he and his team have begun using to study how mutations reshape enzyme activity, contribute to drug resistance and influence hereditary disease. The approach is emerging as a practical alternative to traditional molecular simulations, which often become too computationally heavy to evaluate large mutational landscapes.
Hu, L., Zhang, A., & Warshel, A. (2025). Exploring evolutionary trajectories of drug resistance. Proceedings of the National Academy of Sciences, 122(45). https://doi.org/10.1073/pnas.2517715122
Warshel originally built his reputation on detailed computational models of enzymatic reactions. These methods helped clarify how enzymes accelerate chemical processes inside cells, but they struggled when he attempted to account for the thousands of possible mutations present in real biological systems. Around five years ago, his group began incorporating machine learning and statistical modeling to see if they could capture enzyme behavior without simulating every atomic interaction. They noticed that an enzyme’s overall activity often mirrored a simple measure derived from maximum entropy models. This insight allowed them to estimate how mutations would influence enzyme speed using only statistical information rather than exhaustive physics-based calculations.
Arieh Warshel, a USC stated,
“We keep pushing on drug resistance. But we’re having much more success in enzyme design and in predicting diseases like heart conditions and hearing loss. Maximum entropy works beautifully for these problems.”
Because drug resistance emerges from mutation, the team applied the method to viruses, beginning with HIV. Warshel had previously attempted to model HIV’s evolution using conventional simulations, but even powerful computers struggled to keep up with the virus’s high mutation rate. When the group tested maximum entropy using mutation data from treated HIV patients, the results showed some correlation with known resistance patterns but did not outperform basic descriptors such as the total number of mutations present. HIV’s mutational freedom made it difficult to predict its future paths with meaningful precision.
This limitation prompted the team to explore viruses that mutate in more structured ways. Hepatitis C virus became one of their first case studies, and the outcome was significantly more encouraging. Under drug pressure, HCV tends to follow narrower evolutionary routes, and in this system, maximum entropy closely aligned with the mutations observed in patient samples. This offered the team a way to evaluate not only which mutations were possible, but also which were most likely to appear. Warshel compared the process to playing a slow but strategic game of chess with the virus, where understanding both the strength and the probability of each move becomes essential.
The reach of maximum entropy has since extended far beyond virology. The group applied the method to myosin, a family of molecular motors that enable muscle contraction, hearing and heart function. Mutations in different myosin proteins are known to cause inherited cardiac issues and forms of deafness. Here, too, maximum entropy produced reproducible correlations with disease-causing mutations and with the effects of drugs that modify myosin behavior. A new study from the team, currently under review, suggests that the same statistical approach can even predict differences in how myosin proteins move along cellular tracks when altered by specific mutations.
For engineers and scientists working at the intersection of biology and computation, this shift represents a more efficient way to probe mutation-driven behavior. Maximum entropy avoids the heavy computational burden associated with traditional simulations while still providing quantitative insight into how proteins adapt or fail under different conditions. Although the method has not solved the challenge of forecasting drug resistance in extremely variable viruses like HIV, it has proven effective in systems that follow more constrained evolutionary patterns.
Warshel views the broader lesson as a reminder that streamlined statistical tools can sometimes reveal what complex physics-based models struggle to capture. He continues to pursue drug resistance as a research goal, but much of the group’s progress now lies in enzyme design and in predicting genetic conditions affecting the heart, muscles and hearing. In these areas, maximum entropy has offered clarity with surprising consistency, making it a promising direction for future work in computational biology and bioengineering.
Adrian graduated with a Masters Degree (1st Class Honours) in Chemical Engineering from Chester University along with Harris. His master’s research aimed to develop a standardadised clean water oxygenation transfer procedure to test bubble diffusers that are currently used in the wastewater industry commercial market. He has also undergone placments in both US and China primarely focused within the R&D department and is an associate member of the Institute of Chemical Engineers (IChemE).
