To milliseconds and beyond: challenges in the simulation of protein folding

Abstract

Quantitatively accurate all-atom molecular dynamics (MD) simulations of protein folding have long been considered a holy grail of computational biology. Due to the large system sizes and long timescales involved, such a pursuit was for many years computationally intractable. Further, sufficiently accurate forcefields needed to be developed in order to realistically model folding. This decade, however, saw the first reports of folding simulations describing kinetics on the order of milliseconds, placing many proteins firmly within reach of these methods. Progress in sampling and forcefield accuracy, however, presents a new challenge: how to turn huge MD datasets into scientific understanding. Here, we review recent progress in MD simulation techniques and show how the vast datasets generated by such techniques present new challenges for analysis. We critically discuss the state of the art, including reaction coordinate and Markov state model (MSM) methods, and provide a perspective for the future.

Figure 1

Comparison of predicted and experimentally measured folding times. Central dashed line indicates perfect agreement, outside lines are within one order of magnitude of perfect agreement. Given that experimental folding times can vary over more than an order of magnitude given different conditions (temperature, salt, pH, etc.), as well as uncertainties associated with measuring experimental and simulated folding times, an order of magnitude agreement is close to the upper limit of accuracy one might expect. Data from [,–,,,–92].

Figure 2

The folding times accessible by simulation have increased exponentially over the past decade. Shown are all protein folding simulations conducted using unbiased, all-atom MD in empirical forcefields reported in the literature. Some folding times for the same protein differ, due to various mutations. For lambda marked with a (*), the longest timescale seen in that simulation, which was not the folding time, occurred on the order of 10 ms [18,23]. Data are same as Figure 1, with [93].

Figure 3

Two methods of data analysis, the MSM (top) and reaction coordinate (bottom), shown for the same system (ACBP) [19]. The MSM represents folding as interconversion between structurally similar states, and can be illustrated as flow through a network. Reaction coordinates attempt to depict folding as progress along a single degree of freedom, such as the committors (P_fold, shown). The MSM picture is more detailed, can capture parallel paths, has tunable resolution, and connects naturally to experiment – all advantages over the coordinate-based approach.