laboratoire-de-chemoinformatique/chemspacecopilot/dataset-normalization

Dataset Normalization

Use this skill when uploaded or session-resident tabular data must be prepared for analysis.

Procedure

1. Inspect available data with session_list_loadable_session_data or session_summarize_session_memory.
2. Load existing session data with pandas_load_dataframe_from_session or create a DataFrame with pandas_create_dataframe.
3. Use pandas_run_operation for lightweight cleanup when needed.
4. Normalize the dataset with pandas_normalize_for_analysis.
5. Return the normalized path or DataFrame name for downstream GTM or report workflows.

Expected Outputs

• Analysis-ready normalized dataset.
• Identified SMILES, activity, and optional cluster columns.
• Downstream path or registry name.

Details

• Input resolution order: session GTM source_mols (use its node_index as cluster_id) → clean_dataset_path → dataset_path (legacy clean alias) → ask the user for a path.
• Activity-column detection: raw potency with units (IC50, EC50, Ki, Kd, MIC), p-scale potency (pIC50, pKi, pChEMBL), or labels (activity, label, class). Normalization records activity_mapping (source column) and final_activity_mapping (final merged activity); preserve both for downstream SAR.
• Required column: a SMILES column (smiles / SMILES / canonical_smiles); cluster_id and an activity column are optional but enable clustering and SAR downstream.