RepLikCompare

RepLikCompare is a Python package for working with Molecular Dynamics data: dataset loading, clustering, dimensionality reduction, plotting, and ensemble (replica/system) comparison workflows.

Most functionality is exposed directly on the Dataset class, which wraps a pandas DataFrame carrying simulation/replica metadata and mixes in plotting, clustering, dimensionality-reduction, and ensemble-comparison methods. A dataset can be created from an existing pandas DataFrame or loaded directly from a CSV file:

import RepLikCompare as rlc

# From an existing pandas DataFrame
dataset = rlc.Dataset.from_dataframe(df)

# Or directly from a CSV file
dataset = rlc.Dataset.from_csv("df.csv")

The input dataset must contain two columns identifying, respectively, the system (e.g. WT vs Mutant) and the replica within that system (e.g. 1, 2, 3). See Dataset for auto-detection rules and how to pass these column names explicitly.

Capabilities

Module	Key functions	What it does
Plotting	plot_scatter, plot_free_energy, plot_lineplot_avg, plot_distri_norm, plot_vonmises, plot_rmsd, plot_rmsf, plot_contact_map, plot_cluster_timeline, compute_secondary_structure_timeline	Scatter/line/distribution plots, free-energy landscapes, RMSD/RMSF and contact-map plots, and cluster/secondary-structure timelines – most with a faceted (per-system) variant.
Clustering	compute_cluster_kmean, compute_cluster_GMM, compute_cluster_dbscan, compute_cluster_hdbscan, hierarchical_clustering, assign_cluster_representative	Cluster conformations with K-means, GMM, DBSCAN, HDBSCAN, or hierarchical clustering, and pick a representative frame per cluster.
Dimensionality reduction	compute_pca, compute_umap, compute_tsne, compute_kpca	Reduce a set of numeric features to a low-dimensional embedding (PCA, UMAP, t-SNE, kernel PCA), with optional support for circular/angular features.
Ensemble comparison	compute_jsd, compute_wasserstein, compute_convergence	Statistically compare feature distributions across systems and replicas (Jensen-Shannon divergence, Wasserstein distance) and check trajectory convergence.

Authors

• Alaa REGUEI, PhD Student – Universite Paris Cite, BFA.

• Samuel Murail, Associate Professor – Universite Paris Cite, BFA.