Computational protein design is a transformational field with translational biomedical applications. A model for computational protein design formulates protein design as a search over discrete conformations for all candidate sequences for the minimum-energy conformation, which encodes its corresponding sequence. This formulation is NP-Hard, and thus the design of large proteins remains computationally intractable. One simplifying assumption used by many protein design algorithms ignore long-range interactions between distant residues. These algorithms apply various cutoffs, which eliminate long-range 2-body energies from the energy function. Intuitively, simplifying protein energetics can reduce the complexity of the problem, but must also potentially change the energy, conformation, and even sequence of the computed sequence. I present novel algorithms to exploit the diminishing interactions of long-range residues, and computational experimental data showing the affects of cutoffs on protein design. Notably, I will show that by using provable algorithms, it is possible to reap the efficiency benefits of applying cutoffs to protein design while also recovering the sequence that would be computed without cutoffs.
Jonathan Jou is a PhD candidate in Computer Science at Duke University. He graduated with a B.S. in Computer Science from Duke, and worked for two years as a software development engineer at Microsoft. His interests lie in the intersection between protein design and computer science, particularly in characterizing the challenges of computational structure-based protein design, so that new algorithms can be designed to address these challenges.