Enhanced Sampling with Sub-optimal Collective Variables: Reconciling Accuracy and Convergence Speed

Dhiman Ray,Valerio Rizzi

DOI: https://doi.org/10.26434/chemrxiv-2024-mfpcn

2024-09-18

Abstract:We introduce an enhanced sampling algorithm to obtain converged free energy landscapes of molecular rare events, even when the collective variable (CV) used for biasing is not optimal. Our approach is a combination of the On-the-fly probability enhanced sampling (OPES) and its exploratory variant, OPES Explore (OPESe). We demonstrate the successful application of this combined algorithm on the two-dimensional Wolfe-Quapp potential, ligand-receptor binding in trypsin-benzamidine complex, and folding-unfolding of chignolin. Apart from computing accurate free energy profiles, we can discover additional metastable configurations not distinguished by the sub-optimal CV space. Moreover, we can control the trade-off between accuracy and convergence speed by varying the ratio of the barrier parameters in OPES and OPESe components. The improved efficiency and accuracy of free energy calculation, and the possibility of using generic and intuitive collective variables, make our proposed algorithm particularly promising for the simulation of complex molecular systems.

Chemistry

What problem does this paper attempt to address?

The paper aims to address the issue of insufficient sampling of rare events in molecular dynamics simulations. Specifically, the researchers propose an improved sampling algorithm that can achieve convergent free energy landscapes even when the collective variables (CVs) are not optimal.

### Main Issues:
1. **Insufficient Sampling of Rare Events**: In molecular dynamics simulations, high energy barriers make it difficult for the system to escape local free energy minima, thus preventing adequate sampling of the conformational space. Such rare events are common in many practical applications, such as protein folding, drug-target recognition, and heterogeneous catalysis.
2. **Difficulty in Selecting Collective Variables**: Existing enhanced sampling methods rely on high-quality collective variables to achieve efficient exploration and convergence of the free energy surface. However, finding suitable collective variables is often very challenging, especially when describing complex molecular processes.

### Solution:
The researchers combined the "On-the-fly Probability Enhanced Sampling" (OPES) and its exploratory variant OPES Explore to propose a new algorithm. This method can obtain accurate free energy landscapes even when using suboptimal collective variables and can discover metastable structures that these suboptimal CVs fail to distinguish. Additionally, by adjusting the barrier parameter ratio in the OPES and OPES Explore components, the trade-off between accuracy and convergence speed can be controlled.

### Experimental Validation:
The researchers validated the effectiveness of the algorithm on three different systems:
1. **Wolfe-Quapp Potential Surface**: Tested the algorithm's performance on a 2D potential surface.
2. **Chymotrypsin-Aniline Complex**: Studied the ligand-receptor binding problem.
3. **Chignolin Protein Folding**: Tested the protein folding-unfolding process.

Experimental results show that the proposed algorithm can effectively estimate free energy differences even when using suboptimal CVs and has a faster convergence speed. This makes the algorithm particularly suitable for simulations of complex molecular systems.