Charge-perturbation dynamics - a new avenue towards in silico protein folding

Abstract

Molecular dynamics (MD) has greatly contributed to understanding and predicting the way proteins fold. However, the time-scale and complexity of folding are not accessible via classical MD. Furthermore, efficient folding pipelines involving enhanced MD techniques are not routinely accessible. We aimed to determine whether perturbing the electrostatic component of the MD force field can help expedite folding simulations. We developed charge-perturbation dynamics (CPD), an MD-based simulation approach that involves periodically perturbing the atomic charges to values non-native to the MD force field. CPD obtains suitable sampling via multiple iterations in which a classical MD segment (with native charges) is followed by a very short segment of perturbed MD (using the same force field and conditions, but with non-native charges); subsequently, partially folded intermediates are refined via a longer segment of classical MD. Among the partially folded structures from low-energy regions of the free-energy landscape sampled, the lowest-energy conformer with high root-mean-square deviation to the starting structure and low radius of gyration is defined as the folded structure. Upon benchmark testing, we found that medium-length peptides such as an alanine-based pentadecapeptide, an amyloid-{beta} peptide, and the tryptophan-cage mini-protein can fold starting from their extended linear structure in under 45 ns of CPD (total simulation time), versus over 100 ns of classical MD. CPD not only achieved folding close to the desired conformation but also sampled key intermediates along the folding pathway without prior knowledge of the folding mechanism or final folded structure. Our findings confirmed that perturbing the electrostatic component of the classical MD force field can help expedite folding simulations without changing the MD algorithm or using expensive computing architectures. CPD can be employed to probe the folding dynamics of known, putative, or planned peptides, as well as to improve sampling in more advanced simulations or to guide further experiments.\n\nAuthor summaryFolding represents the process by which proteins assemble into biologically active conformations. While computational techniques such as molecular dynamics (MD) have provided invaluable insight into protein folding, efficient folding pipelines are not routinely accessible. In MD, the behavior of the studied molecule is simulated under the concerted action of multiple forces described by mathematical functions employing optimized parameters. Using non-native parameters effectively perturbs the MD force field. We show that this can be exploited to help expedite folding simulations. Specifically, we developed charge-perturbation dynamics (CPD), an MD-based simulation approach that involves periodically perturbing the force field by using non-native atomic charges. For folding medium-length peptides such as the tryptophan-cage mini-protein starting from the extended linear structure, CPD is much faster than other MD-based approaches while using the same software, hardware, and know-how required for running classical MD simulations. Furthermore, CPD not only achieves folding close to the desired conformation but also samples key intermediates along the folding pathway without prior knowledge of the folding mechanism or final folded structure. CPD can be employed to probe the folding dynamics of known, putative, or planned peptides, as well as to generate different conformations that can guide further experiments or more advanced simulations.