charleshahn/duivyagent

Overview

This skill provides advanced post-processing analysis for protein molecular dynamics (MD) simulations, covering PBC correction, dynamic cross-correlation matrices (DCCM), residue distance contact matrices (RDCM), principal component analysis (PCA), and free energy landscape (FEL) mapping. It guides a reproducible workflow from trajectory cleanup to extraction of low-energy conformations and interpretable visual outputs. The focus is on practical commands, recommended atom groups, and checks to ensure robust results.

How this skill works

The skill inspects MD trajectories and topology to perform periodic boundary condition correction (centering, molecule rejoin, and fit), then computes covariance-based metrics (DCCM, PCA) and distance-based matrices (RDCM). It combines time series (RMSD, radius of gyration or principal components) to build FELs with gmx sham and visualizes XPM outputs using lightweight tools. Each step includes group/atom selection advice, temperature and grid parameter guidance, and methods to extract representative frames.

When to use it

• After completing production MD when trajectories show wrapping or RMSD jumps (perform PBC correction first)
• When you need to quantify correlated motions within the protein (DCCM) or identify residue contacts (RDCM)
• When you want to reduce dimensionality and identify major motion modes (PCA)
• To compute free energy basins and extract minimum-energy conformations (FEL using RMSD/gyrate or PCA axes)
• When preparing figures or metrics for publication, validation, or further ensemble modeling

Best practices

• Always preprocess trajectories: center, ensure molecular integrity, and fit (remove rotation and translation) before any covariance or distance analyses
• Use C-alpha atoms for PCA and DCCM unless side-chain or backbone detail is specifically required
• Analyze only the equilibrated portion of the trajectory (determine by RMSD) to avoid artifacts in covariance and FELs
• Match FEL temperature to the simulation temperature and tune nlevels/grid density to balance resolution and noise
• Visualize intermediate results (trajectory slices, XPM heatmaps, PC projections) to validate automated outputs

Example use cases

• Fix wrapped ligand/protein trajectories and generate a clean fit.xtc for downstream analysis
• Compute a DCCM to reveal regions of correlated and anticorrelated motion and map them onto structure
• Generate an RDCM to find persistent residue-residue contacts and detect allosteric interfaces
• Perform PCA to obtain dominant motion vectors, plot projections, and quantify variance explained
• Build a FEL from RMSD vs gyration or PC1 vs PC2 to locate and extract low-energy conformations