Lightweight Python framework for conformer filtering and deduplication
A lightweight, tool-agnostic Python framework for conformer filtering and deduplication across MD simulations and conformer generation pipelines.
Developed as open-source contribution to Rowan Scientific
Given a large ensemble of molecular conformers from MD simulations, this framework:
- Computes pairwise RMSD matrix (Kabsch alignment)
- Clusters similar conformers using RMSD threshold
- Returns diverse, non-redundant representatives
- Generates diagnostic plots
Tested on Lysozyme (1AKI, 129 residues):
- Input: 50 MD conformers
- Output: ~4-5 diverse representatives
- Redundancy removed: ~88-92%
pip install numpy matplotlib scipy scikit-learn
from conformer_filter import ConformerFilter
cf = ConformerFilter(threshold=2.0)
cf.load_from_arrays(conformers)
cf.fit()
cf.summary()| Method | Description |
|---|---|
| RMSD + Kabsch | Optimal structural superposition |
| Greedy clustering | Fast, interpretable grouping |
| Threshold sensitivity | Benchmarking optimal cutoff |
Yash Singh Sengar — IIT Ropar / NIT Calicut
Open-source contribution to Rowan Scientific