sar-analysis

What it does

Provides a SAR analysis workflow to identify a common scaffold (MCS) across molecules, decompose molecules into a core and R-groups, align fragments to the parent, and produce an interactive HTML report with a designed visualization workflow.

How it works

• Data loading: automatically identifies columns for Compound Key, Activity, and SMILES from the input CSV, rather than assuming fixed names.
• Core identification (MCS): pre-processes molecules with Chem.AddHs, then runs rdFMCS.FindMCS with parameters: threshold=0.8, ringMatchesRingOnly=True, completeRingsOnly=True, atomCompare=rdFMCS.AtomCompare.CompareElements, bondCompare=rdFMCS.BondCompare.CompareOrder. The resulting SMARTS is converted to core_mol, and 2D coordinates are computed with AllChem.Compute2DCoords(core_mol).
• R-group decomposition & refinement: decomposes molecules into core and R-groups; excludes R-group columns that are identical across all molecules and removes those constant points from the core visualization.
• Image generation & alignment: draws molecules using DrawMoleculeACS1996 for consistent visuals. Uses a specific alignment approach to superimpose core and R-groups onto the original molecule by attempting:
  1. direct substructure match,
  2. a match after converting dummies to queries (AdjustQueryProperties(makeDummiesQueries=True)),
  3. a match without chirality (useChirality=False). If a match is found, coordinates are copied from the parent conformer to the fragment; otherwise coordinates are computed anew. The reference implementation for alignment uses:
  • rdRGroupDecomposition.RGroupDecompose([core_mol], mols, asSmiles=False, asRows=False) for matching,
  • helper function align_substructure_to_parent(sub, parent) implementing the three strategies, and
  • synchronization steps to align original and fragment structures.
• HTML Output: generates sar_analysis_report.html with modern CSS styling, ensuring table columns (Original, Core, R-groups) have a minimum width of 300px to preserve readability of structures.

When to use it

Use when you have an activity-bearing molecule set with SMILES and an identifier, and you need to produce an interpretable SAR report with a visually aligned core plus R-groups and an activity heatmap. The workflow is triggered by datasets that fit the described data loading and alignment steps and when a robust MCS core is obtainable with the specified parameters.

What it can touch

• Uses RDKit components: rdFMCS.FindMCS, Chem.AddHs, AllChem.Compute2DCoords, rdMolDraw2D drawing utilities, AdjustQueryProperties, GetSubstructMatch, GenerateDepictionMatching2DStructure.
• Touches file output: writes sar_analysis_report.html.

Caveats

• License is open; no explicit licensing conflict stated beyond described license fields. The workflow relies on RDKit functionality and assumes correct data formatting and column detection.
• The approach assumes the MCS is meaningful (threshold 0.8) and that R-group decomposition yields actionable variable positions; if MCS is too small, the guide recommends clustering first (per the decision framework).