An Open-Source Implementation of the Scaffold Identification and Naming System (SCINS) and Example Applications

Area of Science:

• Computational chemistry
• Cheminformatics
• Drug discovery

Background:

• Organizing chemical structures is crucial for compound library analysis and hit list postprocessing.
• Existing methods like clustering are computationally intensive, while rule-based methods (e.g., Murcko scaffolds) can be too fine-grained.
• There is a need for methods balancing granularity and computational efficiency in chemical space partitioning.

Purpose of the Study:

• To provide an open-source Python implementation of the Scaffold Identification and Naming System (SCINS).
• To characterize SCINS and demonstrate its utility in analyzing chemical databases and guiding compound selection.
• To facilitate wider adoption and application of the SCINS method in cheminformatics.

Main Methods:

• Developed an open-source Python implementation of SCINS, relying solely on the RDKit library.
• Applied SCINS to large chemical databases, including Enamine REAL Diverse and ChEMBL.
• Evaluated SCINS performance in identifying sparse and dense regions of chemical space and grouping bioactive compounds.

Main Results:

• SCINS effectively identifies sparsely and densely populated regions within large chemical datasets.
• Enamine REAL Diverse occupies a smaller SCINS space compared to ChEMBL, contrasting with Murcko scaffold analysis.
• SCINS provides chemically intuitive groupings for medium-sized bioactive compound sets, aiding virtual screening and hit list analysis.

Conclusions:

• The open-source SCINS implementation offers a computationally efficient and balanced approach to chemical structure organization.
• SCINS is valuable for exploring large chemical spaces and selecting relevant compounds for drug discovery.
• This work provides a practical tool and validation for the SCINS methodology in cheminformatics applications.