Unraveling low-resolution structural data of large biomolecules by constructing atomic models with experiment-targeted parallel cascade selection simulations.

Junhui Peng,Zhiyong Zhang

doi:10.1038/srep29360

Junhui Peng, Zhiyong Zhang

Open Access

https://doi.org/10.1038/srep29360

Copy DOI

Abstract

Various low-resolution experimental techniques have gained more and more popularity in obtaining structural information of large biomolecules. In order to interpret the low-resolution structural data properly, one may need to construct an atomic model of the biomolecule by fitting the data using computer simulations. Here we develop, to our knowledge, a new computational tool for such integrative modeling by taking the advantage of an efficient sampling technique called parallel cascade selection (PaCS) simulation. For given low-resolution structural data, this PaCS-Fit method converts it into a scoring function. After an initial simulation starting from a known structure of the biomolecule, the scoring function is used to pick conformations for next cycle of multiple independent simulations. By this iterative screening-after-sampling strategy, the biomolecule may be driven towards a conformation that fits well with the low-resolution data. Our method has been validated using three proteins with small-angle X-ray scattering data and two proteins with electron microscopy data. In all benchmark tests, high-quality atomic models, with generally 1–3 Å from the target structures, are obtained. Since our tool does not need to add any biasing potential in the simulations to deform the structure, any type of low-resolution data can be implemented conveniently.

Highlights

Extra pseudo-energy term based on the given low-resolution data into the molecular mechanics energy function, and the resulting additional forces will deform the biomolecular structure to fit the experimental data
The idea was from a sampling technique called parallel cascade selection molecular dynamics (PaCS-MD)[21,22]
The PaCS-MD method consists of cycles of (1) conformational sampling by multiple independent MD simulations and (2) selecting a number of conformations from the miMD trajectories, which are closest to the target structure[21] or mostly deviate from the average structure[22], to start the cycle of miMD simulations

Summary

Introduction

Extra pseudo-energy term based on the given low-resolution data into the molecular mechanics energy function, and the resulting additional forces will deform the biomolecular structure to fit the experimental data. In our problem of interpreting a given type of low-resolution structural data, we only need to design a scoring function that measures the discrepancy (or similarity) between a simulated conformation and the target experimental data. After each cycle of miMD, we pick a number of conformations with the smallest discrepancy (or highest similarity) to the target data for the cycle. By this iterative screening-after-sampling strategy, our experiment-targeted PaCS-based method (termed as PaCS-Fit) may allow the biomolecule to approach an atomic model that is consistent to the target low-resolution data.

Objectives

Methods

Results

Conclusion