Dr. Vikram K. Mulligan

Seminar Details

Host: Dr. Wenshe Liu

Time: 4:00-5:00pm

Location: BICH Building Rm 108

Seminar Abstract

Computational design methods have permitted the creation of peptides, proteins, and peptidomimetics with applications in medicine, materials science, and manufacturing.  Although recent machine learning advances have enhanced the design of canonical proteins, lack of training data prevents these methods from designing exotic molecules built from synthetic building-bocks, like D-amino acids, beta-amino acids, or peptoid monomers.  Non-canonical design therefore requires treating the design problem as an optimization problem, with a physics-based objective function for scoring candidate sequences.  However, two problems arise: first, protein force fields generalize poorly to exotic chemical building blocks.  Second, the thousands of building blocks available for synthesizing peptides and peptidomimetics create an intractable combinatorial problem.  Ongoing research aims to address both of these problems.  To generalize to more exotic chemical building-blocks, we have developed a bridge permitting any Rosetta protocol to include quantum chemistry calculations, allowing design protocols combining fast, approximate force fields in early stages with general, accurate quantum chemistry methods in later stages.  To tackle the classically-intractable combinatorial problem, we have created the first practical quantum algorithms for molecular design, which now scale to the design of protein-sized molecules on current quantum computers.  Both the quantum chemistry bridge and quantum design algorithms have been implemented as an open-source library, called Masala, which can be linked by Rosetta to permit these methods to be used in any Rosetta design or validation protocol.  These methods should have broad applicability for the development of peptide and peptidomimetic therapeutics, as well as in materials and enzyme engineering.