Abstract
Conformational transitions in proteins can be difficult to observe with equilibrium molecular dynamics and challenging for enhanced sampling methods like Targeted MD when high-resolution structural data are unavailable. Low-resolution data, such as interatomic distances and angles, can serve as collective variables (CVs) to bias steered MD (SMD) simulations, but the optimal choice and number of CVs remain unclear. Here, we identify a minimal set of CVs that drive successful transitions between metastable states in T4 lysozyme. We validate them using temperature-accelerated MD (TAMD) to accelerate conformational changes in the absence of target bias. We found that CVs at both the largest and smallest scales are necessary, including interdomain hinge bending and local side-chain reorientation. A salt bridge between Arg8 and Glu64 stabilizes the closed state and must break for hinge bending, while Phe4 reorients to a hydrophobic pocket to stabilize the open state. Our results highlight the importance of selecting appropriate CVs and optimizing the steering protocol to prevent protein deformation. This work demonstrates that SMD simulations can serve as a predictive tool for understanding protein conformational changes in the absence of high-resolution structural data.