Trajectory-based training enables protein simulations with accurate folding and Boltzmann ensembles in cpu-hours

Abstract

An ongoing challenge in protein chemistry is to identify the underlying interaction energies that capture protein dynamics. The traditional trade-off in biomolecular simulation between accuracy and computational efficiency is predicated on the assumption that detailed force fields are typically well-parameterized, obtaining a significant fraction of possible accuracy. We re-examine this trade-off in the more realistic regime in which parameterization is a greater source of error than the level of detail in the force field. To address parameterization of coarse-grained force fields, we use the contrastive divergence technique from machine learning to train from simulations of 450 proteins. In our procedure, the computational efficiency of the model enables high accuracy through the precise tuning of the Boltzmann ensemble. This method is applied to our recently developed Upside model, where the free energy for side chains is rapidly calculated at every time-step, allowing for a smooth energy landscape without steric rattling of the side chains. After this contrastive divergence training, the model is able to de novo fold proteins up to 100 residues on a single core in days. This improved Upside model provides a starting point both for investigation of folding dynamics and as an inexpensive Bayesian prior for protein physics that can be integrated with additional experimental or bioinformatic data.

Commonly used FRET fluorophores promote collapse of an otherwise disordered protein

“…To investigate whether the conformational ensemble of a realistic heteropolymer results in a significant nonproportionality between R ee and R g , we used Upside, our Cβ-level simulations (54,55), to simulate the scattering for unfolded ensembles of 50 protein of 250-650 residues randomly chosen from the Protein Data Bank (PDB). In its simplest version, Upside represents the polypeptide backbone with six atoms per residue (N, Cα, C, H, O, and Cβ) and uses neighbor-dependent Ramachandran maps derived from a coil library (56).…”

Commonly-used FRET fluorophores promote collapse of an otherwise disordered protein

“…To investigate whether and how decoupling between Ree and Rg alter the SAXS profile and to test our ability to extract information from such deviations, we used Upside, our Cb-level polypeptide chain simulation algorithm (71,72), to simulate the scattering for unfolded ensembles of 50 protein sequences of 250-650 residues randomly chosen from the PDB. In its simplest form, Upside represents the polypeptide backbone with six atoms per residue (N, Ca, C, H, O, Cb) and uses neighbor-dependent Ramachandran maps derived from a coil library (73).…”

“…To test this, we conducted additional simulations using a more detailed version of the Upside algorithm that is capable of de novo folding of proteins shorter than 100 residues (71,72). In this version, each of the 20 side chains is represented by a multi-position eccentric bead that allows for detailed packing of the core.…”

“…Using this model, we generated 30 ensembles for each of six proteins (PNt and five other proteins randomly selected from the list of 50 above), using short simulations that sample only the unfolded state. Ensembles were obtained by running replica exchange simulations over a temperature range from 280-320 K, as described previously (71,72). Values of n obtained from these ensembles ranged from 0.4 to 0.6, depending on the simulation temperature.…”

“…Additionally, our webserver http://sosnick.uchicago.edu/SAXSonIDPs is available for fitting SAXS data with our MFF(n,Rg). Figure S3: Comparison between parameters obtained from the two-parameter MFFs and true values for the conformational ensembles generated using Upside simulations (72). (A-C) Comparison of R g , n, and R ee calculated from coordinates of HP-model simulations versus fit with our published MFF(R g , n) to SAXS profiles with randomly added experimental errors.…”