SigFinder

SigFinder is a tool that exploits statistical significance of substructures from a given compound dataset. It is based on the GraphSig Technology, and can be used for two purposes:

1. Identification of significant substructures
2. Classification of the dataset into categories on the basis of BBB permeability/Toxicity/ADME properties.

Features and Capabilities

• Given a compound dataset, this tool mines statistically significant substructures---substructures that are representative of a given dataset because they are structurally overrepresented or underrepresented. This can help in the design of new drugs with similar capabilities.

• It is capable of mining substructures that would not surface by doing a simple Frequent Substructure search on the dataset. Some of the substructures have less than 0.1% support in our test datasets.
• It can be used as an in silico tool to predict BBB permeability/ Toxicity---properties of molecules that not only depend on pharmacokinetic properties, but the arrangement and interaction of different topological fragments of a molecule.

Method

The workflow of using SigFinder for identifying significant substructures is as follows:

 

The workflow of using SigFinder for classifying chemical compounds is as follows:

 

Validation

• We discovered the following substructures from the ACE inhibitors in the MDDR database.

• We applied SigFinder to discover significant substructures from a dataset of compounds classified as blood-brain barrier permeable and non-permeable. A couple of examples are shown below.

• We applied SigFinder to classify the Anti-cancer screen dataset into actives and inactives. The BEDROC scores comparing our performance to Daylight are shown below.

More details on this technology can be found here.

References

1. Huahai He; Ambuj K. Singh; GraphRank: Statistical Modeling and Mining of Significant Subgraphs in the Feature Space. Proceedings of the 6th IEEE International Conference on Data Mining (ICDM), December, 2006, doi:10.1109/ICDM.2006.79
2. Sayan Ranu; Ambuj Singh; Mining Statistically Significant Molecular Substructures for Efficient Molecular Classification., J. Chem. Inf. Model., 2009, 49(11), pp 2537–2550 DOI : 10.1021/ci900035z