Fully Automated System for Exploring Molecular Dynamics Trajectories
5 affiliations
- 1Department of Physics
- 2California State University Dominguez Hills
- 3Biophysics Program
- 4AAI Research Lab
- 5Department of Biology
Abstract
Molecular dynamics (MD) studies often require separate tools for system preparation, simulation, enhanced sampling, analysis, and reporting, creating opportunities for inconsistent settings and undocumented decisions. We present FastMDXplora, an open-source Python framework built on OpenMM that integrates setup, simulation, analysis, and report phases through a single declarative configuration. A shared schema supports command-line and browser-based graphical interfaces. An optional artificial intelligence (AI) agent generates study configurations from natural-language descriptions, subject to the same validation as user-defined configurations. Each phase records its resolved settings, software versions, and execution environment. When a requested result lacks sufficient support, the software withholds it and reports the reason through a registry of 111 typed refusal codes. Fifteen estimator tests recover analytical benchmarks, including statistical inefficiencies within 2 % of reference values and WHAM reconstruction of analytical free-energy profiles from synthetic umbrella-sampling data with root-mean-square errors below 0.1 kJ mol−1. A labeled corpus of 25 defective and 26 clean studies produces the expected outcomes in all 51 cases. Calibration against ten replicas shows that singlerun standard errors at 20 ns can be up to tenfold smaller than replica-based estimates, even when the estimated correlation time is resolved. The software withholds standard errors when its correlation diagnostics fail, but passing these diagnostics does not establish calibrated uncertainty. In a trypsin–benzamidine case study, interaction analysis reveals differences in the engagement of Asp189 between two ligand charge states that are not distinguished by total contact counts. FastMDXplora provides an integrated framework for reproducible MD studies, combining workflow automation with explicit validation and provenance while preserving the need for independent assessment of sampling and uncertainty. The source code is available at https://github.com/aai-research-lab/FastMDXplora.
Related
- GroScore: Automated Scoring of Protein–Protein Interactions Using Bidirectional Steered Molecular Dynamics
- Metaheuristic Optimization of Temporal Surrogates for Early Prediction of Protein Conformational Stability from Short Molecular-Dynamics Trajectories
- Automated Quenching and Extraction for High-Precision Metabolomics Uncovers Distinct Metabolome Trajectories across Escherichia coli Lineages
- Exploring molecular interactions between yeast topoisomerase I and RNA polymerase III subunit, Rpc82 using an in-silico approach
- RNAactDrug 2.0: an updated resource for systematically exploring molecular features associated with drug sensitivity across multi-omics datasets
BibTeX
@misc{aina2026fully,
title = {{Fully Automated System for Exploring Molecular Dynamics Trajectories}},
author = {Adekunle Aina and Derrick Kwan and Oliche Brown and Obinna Otegbulu and Victory Unegbu and Ayush Kumar},
howpublished = {ChemRxiv},
year = {2026},
month = oct,
doi = {10.26434/chemrxiv.15010333/v1},
url = {https://chemrxiv.org/doi/full/10.26434/chemrxiv.15010333/v1},
note = {CC BY},
}
RIS
TY - UNPB
TI - Fully Automated System for Exploring Molecular Dynamics Trajectories
AU - Adekunle Aina
AU - Derrick Kwan
AU - Oliche Brown
AU - Obinna Otegbulu
AU - Victory Unegbu
AU - Ayush Kumar
PY - 2026
DA - 2026/10/13
PB - ChemRxiv
DO - 10.26434/chemrxiv.15010333/v1
UR - https://chemrxiv.org/doi/full/10.26434/chemrxiv.15010333/v1
AB - Molecular dynamics (MD) studies often require separate tools for system preparation, simulation, enhanced sampling, analysis, and reporting, creating opportunities for inconsistent settings and undocumented decisions. We present FastMDXplora, an open-source Python framework built on OpenMM that integrates setup, simulation, analysis, and report phases through a single declarative configuration. A shared schema supports command-line and browser-based graphical interfaces. An optional artificial intelligence (AI) agent generates study configurations from natural-language descriptions, subject to the same validation as user-defined configurations. Each phase records its resolved settings, software versions, and execution environment. When a requested result lacks sufficient support, the software withholds it and reports the reason through a registry of 111 typed refusal codes. Fifteen estimator tests recover analytical benchmarks, including statistical inefficiencies within 2 % of reference values and WHAM reconstruction of analytical free-energy profiles from synthetic umbrella-sampling data with root-mean-square errors below 0.1 kJ mol−1. A labeled corpus of 25 defective and 26 clean studies produces the expected outcomes in all 51 cases. Calibration against ten replicas shows that singlerun standard errors at 20 ns can be up to tenfold smaller than replica-based estimates, even when the estimated correlation time is resolved. The software withholds standard errors when its correlation diagnostics fail, but passing these diagnostics does not establish calibrated uncertainty. In a trypsin–benzamidine case study, interaction analysis reveals differences in the engagement of Asp189 between two ligand charge states that are not distinguished by total contact counts. FastMDXplora provides an integrated framework for reproducible MD studies, combining workflow automation with explicit validation and provenance while preserving the need for independent assessment of sampling and uncertainty. The source code is available at https://github.com/aai-research-lab/FastMDXplora.
ER -