ADAM10 FAST sampling POPC

This repository contains all the trajectories and input files related to simulations of ADAM10 in a membrane composed of POPC. The xtc trajectories are downsampled 10 times to allow for their publication on zenodo (frames recorded every ns) .As a starting point for the simulations, we used the x-ray structure of ADAM10 (PDB:6BE6). The missing residues (670-748) were reconstructed using Alphafold2 43,44 , resulting in a single transmembrane helix with 2 juxtamembrane helices from residues 654 to 677, that we are calling Stalk Domain. The initial systems were built using CHARMM-GUI 45 . The final ADAM10 model was embedded in a lipid bilayer composed of of POPC. The system was then solvated in a box of explicit water molecules and neutralized in 150 mM KCl. Molecular dynamics simulations of the membrane protein were performed using GROMACS 2023.2. The CHARMM36 force field was employed in combination with the TIP3P water model. The systems underwent energy minimization using the steepest descent algorithm until the energy gradient converged to a threshold of 0.01 kcal/mol/Å. The solvent and membrane were then allowed to relax in the NVT ensemble at 300K for 625ps. The system then underwentNPT equilibration during 3ns, first increasing the timestep from 1fs to 2fs, then progressively removing all constraints applied on protein and lipids. The LINCS algorithm was used for bond constraints. During equilibration, the Berendsen thermostat and barostat (when applicable) were used. For production, the v-rescale thermostat and c-rescale barostat were used for temperature and pressure control, respectively. For equilibration as well as production runs, a cutoff distance fornon-bonded interactions was set at 1.2 nm, utilizing a cut-off van der Waals (vdW) type with a force-switch modifier. The switching distance was configured to 1.0 nm.The Coulombicinteractions were calculated using the Particle Mesh Ewald (PME) method, with a cutoff distance of 1.2 nm.We performed adaptive sampling based on the Fluctuation Amplification of Specific Traits (FAST) method to enhance the exploration of the conformational landscape. Briefly, the method consists of setting up a swarm of short simulations, clustering structures among the resulting trajectories, and selecting new structures among them to set up a new swarm. The new structures were chosen based on a reward function defined as the sum of 127 pairwise distances spread among 4 interfaces between the catalytic domain of ADAM10 and the other subdomains(See publication). We then generated 16 generations of 20 simulations, each with a length of 50 ns. Once the overall sampling done, simulations from all seeds were extended to ensure a better sampling, amounting to 32µs of sampling for each system.