TQchem is the computational engine powering Terra Quantum's drug discovery solution. Built on a unique tensor-network methodology, it enables accurate structure prediction and scoring where conventional AI approaches reach their limits.
A deep dive is a working session with our Life Sciences team. We walk through how TQchem models flexible molecules and binding sites, and the benchmarks behind each method.
Share a target or modality in advance and we will focus the session on where our approach could contribute and how a partnership could be structured.
Orthogonal to conventional ML approaches, TQchem combines tensor-network optimisation with GPU-accelerated quantum chemistry for accurate, scalable molecular simulation.
Quantum Chemistry
GPU-native Hartree-Fock and density functional theory. Up to 10,000 basis functions on a single GPU, scaling across multiple GPUs and nodes.
Conformer Search
Compresses the conformational search space to find low-energy conformers of flexible molecules, including macrocycles and cyclic peptides, at chemical accuracy.
Induced-Fit Docking
Models flexibility in both ligand and binding site. Validated with non-canonical cyclic peptides, a modality that challenges conventional docking.
Tensor networks were first developed to simulate quantum systems too complex to compute directly. We have built them into a proprietary engine for molecular modeling.
It finds the most stable molecular shapes and binding configurations, describing the flexibility of molecule and target together.
faster conformer search than state-of-the-art tools
faster conformer search for large macrocycles versus best-in-class tools
faster quantum chemistry than conventional CPU-based methods
Cyclic peptides with non-canonical residues fall outside what most generative, folding and scoring tools were built for. We applied tensor-train sampling to a virtual screening workflow for a class A receptor, screening a combinatorial library of macrocyclic peptides that varied across multiple chemical dimensions. Ligand pose, backbone and side chains were sampled jointly.
Novel designs scored above the reference while keeping its core pharmacophores. The ranking also surfaced receptor interactions that well-known leads rarely make.
TQchem's core algorithms are documented in peer-reviewed papers, preprints and conference posters spanning conformer search, molecular docking and electronic structure calculation.
Tensor train techniques for conformer search offer a physics-driven approach, unlike many machine learning-based methods that rely on large datasets. Our method finds highly accurate conformers with significantly fewer function evaluations than competing optimisation techniques, delivering substantial speed-ups and a data-independent solution.
We provide free access to selected TQchem components for academic and non-commercial research. Tell us about your project and we will follow up with license details.
We provide free access to selected components for academic and non-commercial research. Apply for a non-commercial license and our team will follow up.
Bring us a flexible molecule, a difficult binding pocket or a system where conventional methods have stalled. In a deep dive, our scientists walk you through how TQchem would approach it.
A deep dive is a working session with our Life Sciences team. We walk through how TQchem models flexible molecules and binding sites, and the benchmarks behind each method.
Share a target or modality in advance and we will focus the session on where our approach could contribute and how a partnership could be structured.
