# Molecular Dynamics

Sparse autoencoders expand the dense 128-dimensional latent space at each ProteinMPNN encoder layer into an overcomplete sparse space, where individual dimensions correlate with protein features such as secondary structure, residue volume, residue shape, disulfide bridges and ligand coordination. On the Brolucizumab short-chain variable fragment, the mod=0, 0.5 and 1 candidates with branched-chain-amino-acid dimensions raised held lower RMSD at 433 K than the native sequence, the ProteinMPNN sequence and the GeoEvoBuilder sequence, while mod=2, the largest raise, unfolded substantially at that temperature; the unmodified mod=0 candidate improved too, showing the autoencoder itself shifts predictions.

Sparse autoencoders expand the dense 128-dimensional latent space at each ProteinMPNN encoder layer into an overcomplete sparse space, where individual dimensions correlate with protein features such as secondary structure, residue volume, residue shape, disulfide bridges and ligand coordination. On the Brolucizumab short-chain variable fragment, the mod=0, 0.5 and 1 candidates with branched-chain-amino-acid dimensions raised held lower RMSD at 433 K than the native sequence, the ProteinMPNN sequence and the GeoEvoBuilder sequence, while mod=2, the largest raise, unfolded substantially at that temperature; the unmodified mod=0 candidate improved too, showing the autoencoder itself shifts predictions.

The structure-based protein language model ESM3 treats proteins as if in vacuum, and without solvent information its per-residue outputs mislabel hydrogen bonds as destabilizing, with error that varies with solvent-accessible surface area. HEIWA adds water–residue attention to ESM3 residue embeddings, passing solvent-mediated force-field information through learnable water key vectors and a rotational angle term. On a thermophile classification proxy task, the HEIWA model's classification metrics stay broadly comparable to the baseline, while per-residue outputs correlate better with an RMSF benchmark and SASA; temperature-ramping molecular dynamics found 23 of 38 tested proteins more stable.

The structure-based protein language model ESM3 treats proteins as if in vacuum, and without solvent information its per-residue outputs mislabel hydrogen bonds as destabilizing, with error that varies with solvent-accessible surface area. HEIWA adds water–residue attention to ESM3 residue embeddings, passing solvent-mediated force-field information through learnable water key vectors and a rotational angle term. On a thermophile classification proxy task, the HEIWA model's classification metrics stay broadly comparable to the baseline, while per-residue outputs correlate better with an RMSF benchmark and SASA; temperature-ramping molecular dynamics found 23 of 38 tested proteins more stable.

KEN embeds two-dimensional convolutional kernels in an E(3)-equivariant graph neural network so that an interatomic potential reads electronic spin-interaction matrices alongside geometric coordinates, sidestepping the ill-posedness of coordinate-only potentials in strongly correlated magnetic systems. On the HE26 actinide dataset, KENJI reaches a training bonding-energy mean absolute error of 6.93 meV/atom against 44.7 meV/atom for MACE-osaka26, and its training force error of 34.10 meV/Å is on par with 26.30 meV/Å. On the independent test set the energy error is 46.74 meV/atom and the force error exceeds MACE-osaka26, which the authors partly attribute to training without stress.

KEN embeds two-dimensional convolutional kernels in an E(3)-equivariant graph neural network so that an interatomic potential reads electronic spin-interaction matrices alongside geometric coordinates, sidestepping the ill-posedness of coordinate-only potentials in strongly correlated magnetic systems. On the HE26 actinide dataset, KENJI reaches a training bonding-energy mean absolute error of 6.93 meV/atom against 44.7 meV/atom for MACE-osaka26, and its training force error of 34.10 meV/Å is on par with 26.30 meV/Å. On the independent test set the energy error is 46.74 meV/atom and the force error exceeds MACE-osaka26, which the authors partly attribute to training without stress.

bAIes originally drew structural information only from AlphaFold2 distograms to build atomic-resolution conformational ensembles of intrinsically disordered proteins. It now accepts six predictors, AlphaFold2, AlphaFold3, Boltz-1, Boltz-2, ESMFold and ESMFold2, and was compared on NMR and SAXS data for 17 disordered proteins. Ensembles from every predictor reach average accuracy comparable to the original AlphaFold2 version, best on local observables such as chemical shifts, RDCs and 3J couplings, and weaker on long-range and global observables such as PREs and SAXS.

bAIes originally drew structural information only from AlphaFold2 distograms to build atomic-resolution conformational ensembles of intrinsically disordered proteins. It now accepts six predictors, AlphaFold2, AlphaFold3, Boltz-1, Boltz-2, ESMFold and ESMFold2, and was compared on NMR and SAXS data for 17 disordered proteins. Ensembles from every predictor reach average accuracy comparable to the original AlphaFold2 version, best on local observables such as chemical shifts, RDCs and 3J couplings, and weaker on long-range and global observables such as PREs and SAXS.

AI agents can now propose atom-level biological hypotheses faster than they can be tested, but trust in an agent-run molecular dynamics simulation rests on a process record, and existing agents leave that record to the language model itself. MDClaw pairs validating tools with a directed acyclic graph (DAG, a graph recording which calculation stages depend on which) that checks stages before execution and records each one; on MDDataBench, which grades against deposited simulations, four open-weight models with MDClaw passed all three runs on 81–100% of 31 test systems. An ablation without the DAG found 11 runs that passed every output check yet had skipped minimization or equilibration, or simulated a different structure.

AI agents can now propose atom-level biological hypotheses faster than they can be tested, but trust in an agent-run molecular dynamics simulation rests on a process record, and existing agents leave that record to the language model itself. MDClaw pairs validating tools with a directed acyclic graph (DAG, a graph recording which calculation stages depend on which) that checks stages before execution and records each one; on MDDataBench, which grades against deposited simulations, four open-weight models with MDClaw passed all three runs on 81–100% of 31 test systems. An ablation without the DAG found 11 runs that passed every output check yet had skipped minimization or equilibration, or simulated a different structure.

Coarse-grained molecular dynamics gives short membrane-protein trajectories, but long-time diffusion statistics are costly to obtain. A neural stochastic model combines a transformer encoding of protein configurations, recurrent memory at multiple time scales and a stochastic latent process with learned memory to generate correlated displacements autoregressively. In the seven-PIEZO system it yields a mean anomalous diffusion exponent of 1.048±0.012 against 1.049±0.022 for MD, where the earlier VAR-SDE model gives 1.563±0.016, and its trajectories extend to 40,000 frames.

Coarse-grained molecular dynamics gives short membrane-protein trajectories, but long-time diffusion statistics are costly to obtain. A neural stochastic model combines a transformer encoding of protein configurations, recurrent memory at multiple time scales and a stochastic latent process with learned memory to generate correlated displacements autoregressively. In the seven-PIEZO system it yields a mean anomalous diffusion exponent of 1.048±0.012 against 1.049±0.022 for MD, where the earlier VAR-SDE model gives 1.563±0.016, and its trajectories extend to 40,000 frames.

Cubic PbTe keeps a centrosymmetric average structure yet hosts finite-range local polar correlations, which were often attributed to lattice anharmonicity. Machine-learned interatomic potentials trained on density functional theory data drive 100,000-atom-scale molecular dynamics, and mapping them onto an interpretable force-constant model lets third- and fourth-order terms be switched off. The nanoscale correlations persist without those anharmonic terms, but vanish when harmonic interactions are cut to the third nearest neighbor; at low temperature the correlation range exceeds 50 Å.

Cubic PbTe keeps a centrosymmetric average structure yet hosts finite-range local polar correlations, which were often attributed to lattice anharmonicity. Machine-learned interatomic potentials trained on density functional theory data drive 100,000-atom-scale molecular dynamics, and mapping them onto an interpretable force-constant model lets third- and fourth-order terms be switched off. The nanoscale correlations persist without those anharmonic terms, but vanish when harmonic interactions are cut to the third nearest neighbor; at low temperature the correlation range exceeds 50 Å.