CG simulations reproduce the correct lipid-binding pose to LTPs. (A) Unbiased CG-MD simulation protocol used in this study. Atomistic structures of the protein were coarse-grained and set up with a lipid randomly placed in a bulk solvent far away from the protein, followed by MD simulations. The lipid tail solvation (water beads within 5.0 Å of the lipid tails, represented as blue spheres) decreases as the lipid moves from bulk solvent to the hydrophobic binding cavity. (B) Lipid tail solvation in the last 100 ns of the CG-MD trajectories. Light purple indicates true positives, gray indicates false negatives, and dark purple indicates true positives after refinement of the simulation parameters. The two dotted lines indicate the average lipid tail solvation for the true positives (purple, 8/13 proteins) and the false negatives (gray, 5/13 proteins). (C) Distance between the center of mass (COM) of the hydrophobic tails of the experimental ligands and the lipids tested in the simulations. Only the frames with a lipid tail solvation <2 were considered. (D) Illustrative examples of agreement between lipid positions arising from our protocol (lipid tails: yellow, headgroups: purple) and the lipid/ligand position in x-ray structures (orange). The use of a different force constant for the elastic network of StARD11 is indicated by a *. Plots in B and C contain box plots and kernel density estimates (violin plots) of lipid tail solvation and distance, respectively. Box plots show the interquartile range as a white box, containing a horizontal line that depicts the median as well as minimum and maximum values shown as whiskers, which match the range displayed in the violin plots. Outliers are represented as white circles.