Abstract
Understanding how lipids interact with lipid transfer proteins (LTPs) is essential for uncovering their molecular mechanisms. Yet, many available LTP structures, particularly those thought to function as membrane bridges, lack detailed information on where their native lipid ligands are located. Computational strategies, such as docking or AI-methods, offer a valuable alternative to overcome this gap, but their effectiveness is often restricted by the inherent flexibility of lipid molecules and the lack of large training sets with structures of proteins bound to lipids. To tackle this issue, we introduce a reproducible computational pipeline that uses unbiased coarse-grained molecular dynamics (CG-MD) simulations on a free and open-source software (GROMACS) with the Martini 3 force-field. Starting from a configuration of a lipid in bulk solvent, we run CG-MD simulations and observe spontaneous binding of the lipid to the protein. We show that this protocol reliably identifies lipid-binding pockets in LTPs and, unlike docking methods, suggests potential entry routes for lipid molecules with no a priori knowledge other than the protein's structure. We demonstrate the utility of this approach in investigating bridge LTPs whose internal lipid-binding positions remain unresolved. Altogether, our study provides a cost-effective, efficient, and accurate framework for mapping binding sites and entry pathways in diverse LTPs. Key features • Demonstrates the reliability of unbiased coarse-grain molecular dynamics (CG-MD) simulations with the Martini 3 force-field in identifying lipid-binding sites in lipid transfer proteins (LTPs). • The protocol is straightforward to replicate, relying solely on freely available open-source software. • Furthermore, it is computationally efficient, with most simulations completing within a few hours on a standalone GPU-accelerated workstation. • As input, the user only needs to include the structure of the protein and select the lipid type to test.