Tensorizing Pairwise Structural Analysis: GPU-Accelerated RMSD and FCC Clustering in HADDOCK3 Without Structural Regression.
- 1University of Ghana
Abstract
Integrative macromolecular docking pipelines generate thousands of candidate structural decoys, relying on downstream pairwise ensemble analysis, including Root-Mean-Square Deviation (RMSD) matrices and Fraction of Common Contacts (FCC) clustering, to identify biologically relevant binding poses. However, pairwise structural comparisons scale quadratically (O(N^2)), and canonical workflows in HADDOCK3 historically rely on disk-bound trajectory serialization and iterative CPU set operations, creating a quadratic computational and storage bottleneck for large ensembles. Here, we demonstrate that these operations can be reformulated as continuous linear algebra on GPU-resident tensors: structural superposition via batched Singular Value Decomposition (SVD) and residue contact overlap via binary matrix multiplication (M = A A^T). Across empirical benchmarks spanning N = 200 to 10,000 models (evaluating up to 49,995,000 decoy pairs), our GPU implementation achieves a 6.41-fold compute speedup on Apple Silicon (MPS) and up to 12.2-fold end-to-end acceleration in an N = 5,000 ablation on NVIDIA CUDA by eliminating trajectory disk I/O. Crucially, the implementation achieves zero downstream structural regression relative to canonical baselines, preserving identical cluster assignments (Adjusted Rand Index ARI = 1.0000, zero differing assignments) with sub-microangstrom RMSD concordance (Delta_max <= 4.07 x 10^-7 Angstrom on single-precision MPS; machine-precision Delta_max <= 8.88 x 10^-16 Angstrom on double-precision CUDA). The accelerated code, test suites, and benchmarking workflows are openly accessible on GitHub under the Apache-2.0 license.
Related
- Structural clustering across the tree of life reveals candidate Asgard-eukaryote homologs and a shared expansion of solenoids
- Structural Evolution of Neural Networks during Zebrafish Early Development
- Structural Determinants of Ligand-Mediated Hmg2 Mallostery
- Edge Effects on Carbon, Nitrogen, Phosphorus, and Non-Structural Carbohydrates in Organs of Salsola collina
- Genomic structural variations distinguishing archaic and modern humans are associated with neurodevelopment
BibTeX
@misc{salami2026tensorizing,
title = {{Tensorizing Pairwise Structural Analysis: GPU-Accelerated RMSD and FCC Clustering in HADDOCK3 Without Structural Regression.}},
author = {R. O. Salami and S. M. Kumordzi and D. N. Amuna and R. A. Carr and C. M. Morang'a},
howpublished = {bioRxiv},
year = {2026},
month = oct,
doi = {10.64898/2026.10.10.758202},
url = {https://www.biorxiv.org/content/10.64898/2026.10.10.758202v1},
note = {CC BY-NC},
}
RIS
TY - UNPB
TI - Tensorizing Pairwise Structural Analysis: GPU-Accelerated RMSD and FCC Clustering in HADDOCK3 Without Structural Regression.
AU - R. O. Salami
AU - S. M. Kumordzi
AU - D. N. Amuna
AU - R. A. Carr
AU - C. M. Morang'a
PY - 2026
DA - 2026/10/10
PB - bioRxiv
DO - 10.64898/2026.10.10.758202
UR - https://www.biorxiv.org/content/10.64898/2026.10.10.758202v1
AB - Integrative macromolecular docking pipelines generate thousands of candidate structural decoys, relying on downstream pairwise ensemble analysis, including Root-Mean-Square Deviation (RMSD) matrices and Fraction of Common Contacts (FCC) clustering, to identify biologically relevant binding poses. However, pairwise structural comparisons scale quadratically (O(N^2)), and canonical workflows in HADDOCK3 historically rely on disk-bound trajectory serialization and iterative CPU set operations, creating a quadratic computational and storage bottleneck for large ensembles. Here, we demonstrate that these operations can be reformulated as continuous linear algebra on GPU-resident tensors: structural superposition via batched Singular Value Decomposition (SVD) and residue contact overlap via binary matrix multiplication (M = A A^T). Across empirical benchmarks spanning N = 200 to 10,000 models (evaluating up to 49,995,000 decoy pairs), our GPU implementation achieves a 6.41-fold compute speedup on Apple Silicon (MPS) and up to 12.2-fold end-to-end acceleration in an N = 5,000 ablation on NVIDIA CUDA by eliminating trajectory disk I/O. Crucially, the implementation achieves zero downstream structural regression relative to canonical baselines, preserving identical cluster assignments (Adjusted Rand Index ARI = 1.0000, zero differing assignments) with sub-microangstrom RMSD concordance (Delta_max <= 4.07 x 10^-7 Angstrom on single-precision MPS; machine-precision Delta_max <= 8.88 x 10^-16 Angstrom on double-precision CUDA). The accelerated code, test suites, and benchmarking workflows are openly accessible on GitHub under the Apache-2.0 license.
ER -