DynoDB: Small Protein Conformational Ensemble Dataset

Welcome to DynoDB, a comprehensive dataset of small protein conformational ensembles generated from all-atom molecular dynamics (MD) simulations. DynoDB focuses on small proteins (5–100 amino acids), capturing their high conformational flexibility, functional diversity, and dynamic properties.

This repository contains all the scripts, notebooks, and workflows used to generate, analyze, validate, and annotate the DynoDB dataset.

---

Dataset Overview

• Proteins: 8,385 small proteins
• Simulation length: 100 ns per protein (all-atom MD)
• Trajectory data: 9.1 TB of XTC trajectories
• Annotations: Structural and functional information
• Analyses: MD trajectory analysis, PCA, Jensen-Shannon distance, secondary structure, Ramachandran plots

---

Useful Links

• Dataset on Hugging Face

---

Repository Structure

DynoDB/
├── MD_Simulation_Workflow/               # Scripts to setup and run MD simulations (GROMACS)
│   ├── 1A11.pdb
│   ├── ions.mdp
│   ├── md.mdp
│   ├── minim.mdp
│   ├── npt.mdp
│   ├── nvt.mdp
│   └── README.md
├── Functional_Annotation/                # Protein functional annotation scripts
│   ├── annotate_proteins.py
│   ├── 8385_ID.xlsx
│   └── README.md
├── MD_Pipeline_Reproducibility_Analysis/ # Workflow to validate MD reproducibility (PCA, JS distances)
│   ├── run_pca_js_batch.py
│   └── README.md
├── MD_Trajectory_Analysis_Workflow/      # Jupyter notebook for in-depth trajectory analysis
│   ├── 1A11
│   ├── MD_Trajectory_Analysis_Workflow.ipynb
│   └── README.md
└── README.md                             # This file

---

Modules Overview

1. MD Simulation Workflow

• Provides step-by-step guidance to generate MD trajectories for small proteins using GROMACS 2024.1.
• Covers: topology generation, solvation, ion addition, energy minimization, NVT/NPT equilibration, and production runs.

2. Protein Functional Annotation

• Python script to classify proteins into 17 functional categories (e.g., antimicrobial, membrane-associated, toxin modulators) using UniProt queries and keyword matching.
• Outputs an Excel file with TRUE/FALSE flags and matched keywords.
• Supports caching via uniprot_cache.json to speed up repeated runs.

3. MD Pipeline Reproducibility Analysis

• Validates MD reproducibility by comparing our trajectories (ours) with benchmark datasets (e.g., ATLAS).
• Computes pairwise Cα distances, performs PCA, and calculates Jensen-Shannon distances to quantify structural similarity.
• Generates plots and heatmaps for all proteins and replicates.

4. MD Trajectory Analysis Workflow

• A Jupyter Notebook for end-to-end MD trajectory analysis.
• Includes preprocessing, RMSD/RMSF, radius of gyration (Rg), SASA, PCA, clustering, DSSP secondary structure, and Ramachandran plot analysis.
• Designed as a template for small protein analysis.

---

Usage

1. Run MD simulations: Follow MD_Simulation_Workflow/README.md.
2. Annotate protein functions: Run Functional_Annotation/annotate_proteins.py.
3. Validate pipeline reproducibility: Run MD_Pipeline_Reproducibility_Analysis/run_pca_js_batch.py.
4. Analyze MD trajectories: Open MD_Trajectory_Analysis_Workflow/MD_Trajectory_Analysis_Workflow.ipynb and execute cells sequentially.

Dependencies:

• Python packages: numpy, pandas, matplotlib, seaborn, scikit-learn, mdtraj, biopython, scipy, requests, psutil
• GROMACS ≥ 2024.1 (GPU optional)
• Internet connection required for UniProt queries

---

Citation