Molecular Dynamics

Molecular Dynamics

Molecular Dynamics

Aug 6, 2025

|

9

min read

The Engine

The Molecular Dynamics Platform runs end to end simulations and analyses for proteins, protein–ligand complexes, and ligand–membrane systems. It automates structure preparation, solvation, equilibration, and production runs, then layers binding energy and motion analysis to translate trajectories into decisions. Core tools include Protein in Water MD with GROMACS, Protein–Ligand MD with optional umbrella sampling and MMPBSA, Ligand Membrane Simulation with enhanced sampling, and a Protein Preparation utility for structure cleanup. Boltz augments these workflows by predicting fold, complexes, and pocket specific binding to prioritize systems before and after MD.

The Algorithm

The platform standardizes MD setup, then tailors engines to the question at hand.

  • Protein Preparation, removal of waters and heteroatoms if desired, missing atom rebuild, hydrogen and disulfide assignment, pH aware protonation, chain selection, interactive 3D review.

  • Protein in Water MD, AMBER99SB-ILDN force field with TIP3P, TIP4P-EW, or SPC/E water, cubic, dodecahedron, or octahedral boxes, user defined edge distance, automated minimization, NVT, NPT, and production. Optional PCA maps dominant motions.

  • Protein–Ligand MD, Amber14 parameters with TIP3P water, standard cutoffs at 1.2 nm, H bond constraints, trajectory analytics for RMSD, RMSF, radius of gyration, and optional MMPBSA or MMGBSA with per residue decomposition. Umbrella sampling produces PMF profiles after steered MD seeding.

  • Ligand Membrane Simulation, GROMACS with CHARMM, AMBER, or GROMOS compatibility, 200 plus lipid types and mixed bilayers, standard MD or enhanced sampling using umbrella sampling, steered MD, or well tempered metadynamics via PLUMED.

  • Boltz AI Prediction, pipelines for single proteins, multimers, pockets, and affinity, outputs pTM, ipTM, and pIC50 like scores to triage systems and seed MD starting structures.

All jobs are created as named pipelines, progress is tracked in Command Center, inputs can be uploaded or passed from Data Engineering, and results include trajectories, energies, plots, and downloadable tables.

Algorithm Validation

Workflows mirror community standards for biomolecular MD preparation and analysis. Stability checks monitor energy conservation, temperature, pressure, and density across equilibration and production. Structural metrics are interpreted with conventional guidance, RMSD under 2 Å indicates stable backbones, 2 to 4 Å indicates moderate flexibility, above 4 Å suggests instability or major transitions. Binding free energies from MMPBSA or MMGBSA are reported with component terms and frame statistics, ΔG_bind below about −5 kcal mol indicates strong binding, window overlap and PMF smoothness are required for umbrella sampling acceptance. PCA reproducibility, convergence of time averaged properties, and agreement with known literature ranges are used as acceptance criteria before decision making.

Scientific Impact

The platform connects preparation, sampling, and inference to answer mechanism and potency questions. It reveals conformational ensembles that inform ensemble docking and induced fit hypotheses, quantifies residue wise binding contributions that guide analog design, characterizes membrane permeation barriers for ADMET assessment, and projects dominant collective motions that explain activity cliffs and cryptic pocket exposure. Boltz predictions close the loop, ranking folds, complexes, and pockets to choose what to simulate next, while MD validates and refines AI hypotheses.

Business Impact

By unifying preparation, simulation, enhanced sampling, and AI analysis in one place, the Molecular Dynamics Platform shortens iteration cycles, reduces setup errors, and concentrates compute where it matters. Teams gain faster target triage, clearer no go decisions, and higher confidence in go decisions, fewer failed synthesis rounds due to unstable poses, quantitative evidence for portfolio reviews, and reusable, auditable pipelines that scale from quick 0.1 ns validations to production campaigns with binding free energy endpoints.