For more software details, check out the group GitHub at: https://github.com/THGLab/
ECENet. ECENet is an equivariant machine learning interatomic potential built around edge-centered representations. It uses persistent SO(2)-equivariant features on atom-pair edges and equivariant message passing to predict energies, forces, and stress, with an ASE interface for molecular dynamics and structural relaxation. The associated Edge Cluster Expansion architecture is described in a 2026 preprint. GitHub
NewtonNet. NewtonNet is a Newtonian message-passing neural network for learning interatomic energies and forces. Recent work extends the differentiable model to obtain analytical Hessians for transition-state optimization, substantially reducing the cost of Hessian evaluation compared with direct ab initio calculations. GitHub
torchff. torchff is a high-performance differentiable molecular mechanics force-field library implemented in PyTorch with custom CUDA kernels. It provides differentiable implementations of common energy terms, including bonds, angles, torsions, electrostatics, and van der Waals interactions, while retaining PyTorch automatic differentiation. GitHub
SmileyLlama. SmileyLlama adapts Llama models for directed exploration of chemical space. The repository contains the code and data used to train and evaluate SmileyLlama and its variants, including models for property-conditioned molecular generation and generation of candidate binders. GitHub
SynLlama. SynLlama is a fine-tuned Llama model for generating synthesizable molecules and molecular analogs through complete synthetic pathways based on accessible building blocks and reaction templates. The repository provides inference and retraining code, with trained models and related resources available separately. GitHub
LinkLlama. LinkLlama is a Llama-based model for chemically reasonable linker design between molecular fragments using geometric and molecular-property constraints. In addition to the software, the project releases pretrained model weights, training data, benchmark sets, and processed ChEMBL data. GitHub
CAFE. Cofolding-Approach to Fragment Exploration is a workflow for exploring orthosteric, allosteric, and cryptic binding sites with fragment libraries and protein–ligand cofolding. The repository contains fragment catalogs, pocket reference structures, and the workflow used for cofolding and binding-site localization. GitHub
KinConfBench. KinConfBench is a reproducible benchmark for evaluating protein–ligand cofolding models on kinase conformational states. The repository provides workflows for running several cofolding models, assigning kinase conformational states, and analyzing model predictions, together with released inference data. GitHub
IDPForge. IDPForge is a transformer protein-language diffusion model for generating all-atom ensembles of intrinsically disordered proteins and disordered regions while retaining accompanying folded domains. The repository provides the model implementation and workflows for ensemble generation. GitHub
LMHE-TSopt. LMHE-TSopt is a machine-learning-driven transition-state optimizer that directly predicts the leftmost Hessian eigenvector (LMHE), the mode that locally approximates the reaction coordinate, rather than requiring a full Hessian calculation at every optimization step. The method uses an E(3)-equivariant neural network and integrates the predicted LMHE into restricted-step partitioned rational function optimization (RS-PRFO). GitHub
X-EISDv2. X-EISDv2 is a generalized, modular, and faster implementation of Extended Experimental Inferential Structure Determination for evaluating structural ensembles of intrinsically disordered proteins against experimental data. It updates the original X-EISD implementation with a more modular standalone software package. GitHub
UCBShift 2.0. UCBShift 2.0 extends UCBShift from backbone chemical-shift prediction to both backbone and side-chain atoms, enabling whole-protein chemical-shift analysis. The method combines sequence- and structure-based transfer prediction with an ensemble decision-tree machine-learning model using structural features derived from protein structures. The GitHub repository contains the software, training and testing identifiers, data-preparation scripts, and procedures for reproducing and retraining the model; the trained UCBShift 2.0 model files are available through Zenodo. GitHub
Hydrogen Combustion. A Benchmark Data Set for Hydrogen Combustion (2021). Data that was developed for machine learning applications deposited here: GitHub
UCBShift: Accurate Prediction of Chemical Shifts for Aqueous Protein Structure on “Real World” Data (2020). Jie Li, Kochise C. Bennett, Yuchen Liu, Martin V. Martin, and T. Head-Gordon. Chem. Sci. 11, 3180-3191 GitHub
MB-UCB: A. K. Das, L. Urban, I. Leven, M. Loipersberger, A. Aldossary, M. Head-Gordon, T. Head-Gordon (2019). Development of an Advanced Force Field for Water using Variational Energy Decomposition Analysis. J. Chem. Theory Comput. 15 (9), 5001-5013 [link] GitHub
MR-3D-DenseNet. S. Liu, J. Li, K. C. Bennett, B. Ganoe, T. Stauch, M. Head-Gordon, A. Hexemer, D. Ushizima, and T. Head-Gordon (2019) Multiresolution 3D-DenseNet for Chemical Shift Prediction in NMR Crystallography. J. Phys. Chem. Lett. 10, 4558−4565. DOI: 10.1021/acs.jpclett.9b01570. Github
PB-AM and PB-SAM. E. Jurrus, D. Engel, K. Star, K. Monson, J. Brandi, L. E. Felberg, D.H. Brookes, L. Wilson, J. Chen, K. Liles, M. Chun, P. Li, T. Dolinsky, R. Konecny, D. Koes, J. E. Nielsen, T. Head-Gordon, W. Geng, R. Krasny, M. Gunner, G.-W. Wei, M. J. Holst, J. A. McCammon, N. A. Baker (2018). Improvements to the APBS biomolecular solvation software suite. Protein Sci 27 (1), 112-128. doi: 10.1016/j.str.2014.10.011 ; Github
Berkeley-SC-Ensemble Decoys. A. Bhowmick, T. Head-Gordon (2015). A monte carlo method for generating side chain structural ensembles. Structure 23(1):44-55. doi: 10.1016/j.str.2014.10.011 ; M. S. Lin, N. L. Fawzi, and T. Head-Gordon (2007). Hydrophobic potential of mean force as a solvation function for protein structure prediction. Structure 15, 727-740 (2007). http://DOI:10.1016/j.str.2007.05.004. Berkeley-SC-Ensemble
Temperature Cool Walking. J. Lincoff, S. Sasmal, and T. Head-Gordon (2016). Comparing generalized ensemble methods for sampling of systems with many degrees of freedom. J. Chem. Phys. 145(17), 174107.http://doi.org/10.1063/1.4965439 S. Brown & T. Head-Gordon (2003). Cool-walking: a new markov chain monte carlo sampling method. J. Comp. Chem. 24 (1), 68-76 (PAK Symposium).http://doi/abs/10.1002/jcc.10181
The code here is a top-level python script that can run TCW simulations using the OpenMM software package. https://github.com/thglab/cool-walking