datamol

What it does

Datamol provides a Pythonic abstraction over RDKit to simplify molecular cheminformatics tasks. It supports molecular format conversions (SMILES, SELFIES, InChI), structure standardization and sanitization, descriptor and fingerprint calculations, 3D conformer generation, clustering and diversity selection, scaffold/fragment analysis, chemical reactions, visualization, and batch processing with parallelization. All molecular objects are native rdkit.Chem.Mol instances.

How it works

The skill instructs the agent to perform concrete steps such as:

• Create molecules from SMILES using a function to_mol, and handle errors by returning None when parsing fails.
• Convert molecules to SMILES, including canonical and isomeric variants, and obtain InChI, InChIKey, and SELFIES.
• Sanitize and standardize molecules, either individually or in bulk, with options like disconnect_metals, normalize, and reionize; also standardize SMILES strings directly.
• Read and write molecular data from various formats (SDF, SMILES, CSV, Excel, and universal readers), including remote paths, using specified functions like read_sdf, read_smi, read_csv, read_excel, and open_df; similarly write with to_sdf, to_smi, to_xlsx.
• Compute descriptors (single and batch) and filter by drug-likeness criteria (MW, logP, HBD, HBA) based on computed descriptors.
• Generate and compare fingerprints (ecfp and others), and compute pairwise distances or similarities between molecules.
• Cluster molecules (Butina) and perform diversity selection; support centroids extraction and capacity to warn about full distance matrix requirements.
• Extract Murcko scaffolds, group by scaffold, and perform scaffold-based train/test splitting for ML tasks.
• Fragment molecules with BRICS and RECAP, and analyze fragments across libraries.
• Generate 3D conformers, cluster conformers by RMSD, and access SASA values from conformer properties.
• Visualize molecules and aligned SAR analyses, highlight substructures, and render conformer grids.
• Apply chemical reactions using RDKit’s reaction framework, including batch application across libraries.
• Utilize built-in parallelization by leveraging n_jobs settings for supported functions (e.g., read_sdf, batch descriptor computation, clustering, distance calculations, SASA).

When to use it

• When standard drug-discovery workflows require straightforward, parallelizable cheminformatics tasks with RDKit compatibility.
• When you need a unified, high-level API for conversions, standardization, descriptors, fingerprints, 3D conformers, clustering, scaffolds, and visualizations without writing RDKit glue code.

What it can touch

• Import convention: import datamol as dm
• Core operations touch molecules (Mol objects) and DataFrames through provided read/write APIs, descriptor pipelines, clustering, and visualization utilities. Parallelization is supported via n_jobs in multiple functions.

Caveats

• Returns native rdkit.Chem.Mol objects for compatibility with RDKit ecosystems.
• For advanced control or custom parameters, use rdkit directly.
• Some statements reference capabilities like cloud storage via fsspec and progress bars on batch operations; these are dependent on the respective functions being invoked with proper parameters.