Simulation Guided Design of Spectroscopy Experiments via Maximizing Kinetic Information Gain

Abstract

We present a computational tool called Optimal Probes that uses protein molecular dynamics (MD) simulation datasets to address a systematic issue in designing spectroscopic experiments. A common problem faced in experiments that measure relaxation of a pair of residues is narrowing down the set of residue pairs to site-specifically label among a large pool of possibilities. Our method ranks every possibility based on a quantitative measure, Generalized Matrix Rayleigh Quotient (GMRQ) score for a Markov state model (MSM) of protein conformational dynamics[1]. We optimize the choice of residue pairs in the protein to build an MSM that can achieve a high GMRQ score, thus capturing slow motions of the protein's dynamics. Once all residue pair sets are scored, the highest scoring inter-residue pairs can be probed to investigate the protein experimentally. We show that incorporating protein dynamics from simulations in such a manner maximizes information gain from experiments. We illustrate that Optimal Probes can make better choices than experimental choices based on human intuition or limited structural information such as distance difference between two known stable structures of the protein. The tool can also be used to predict optimal experiment choices considering previous experiments that have already been performed on the protein of interest. We demonstrate the use of Optimal Probes as an intuitive and user-friendly tool on cytoplasmic and transmembrane proteins for four experimental techniques, Double Electron-Electron Resonance[2], Trp-Tyr quenching, Luminescence Resonance Energy Transfer, and Triplet-Triplet Energy Transfer. In conclusion, Optimal Probes provides an automated prediction of residue pairs for biophysical experiment techniques. [1] Mittal and Shukla. J Phys Chem B, 2017 DOI: 10.1021/acs.jpcb.7b04785 [2] Selvam, Mittal, and Shukla. ACS Central Science, 2018 DOI: 10.1021/acscentsci.8b00330.