An IRG2 team, led by Prof. Sahar Sharifzadeh (Boston University) and Prof. Stacy Copp (University of California, Irvine), has developed an approach that integrates machine learning with density functional theory (DFT) calculations to study dynamic electronic structure fluctuations in the de novo peptide ACC-Dimer. In biomolecular environments, it remains unclear which chemical and structural dynamics support electronic conductivity, which is an essential factor for designing charge-conducting interfaces between biological systems and electronic devices.

To address this, the team examined how finite-temperature fluctuations affect electronic structure and conductivity in ACC-Hex, a peptide-based fiber material composed of an antiparallel coiled-coil hexamer. By combining all-atom classical molecular dynamics (MD) with first-principles DFT and interpretable machine learning, they analyzed the relationship between the physical and electronic structure of the peptide dimer subunit of ACC-Hex. Hybrid DFT calculations for thousands of unique MD snapshots revealed significant variations in near-gap orbital energies. Notably, they observed an increase in the number of nearly degenerate states near the highest occupied molecular orbital (HOMO), suggesting enhanced conductivity. Interpretable machine learning was then used to identify key nuclear conformations contributing to this effect. As expected, interphenylalanine distance and orientation were highly correlated with increased state density near the HOMO. Surprisingly, the analysis also highlighted the importance of tightly coiled peptide backbones and electrostatic environment changes in predicting the number of hole-accessible states.

This study demonstrates the power of interpretable machine learning in uncovering complex trends in large-scale ab initio simulations, offering valuable insights into peptide-based electronic materials. Learn More

Mastracco, P.; Mohanam, L. M.; Nagaro, G.; Prusty, S.; Oh, Y.; Wu, R.; Cui, Q.; Hochbaum, A. I.; Copp, S. M.; Sharifzadeh, S. Journal of Chemical Information and Modeling 2025 65 (5), 2503-2512.

This work has been published in Computational Chemistry. The research team included collaborators from CCAM Stacy Copp, Ruqian. Wu, Allon Hochbaum and Sahar Sharifzadeh.