Related Experiment Videos
Classification of biologically active compounds by median partitioning
Jeffrey W Godden1, Ling Xue, Jürgen Bajorath
1Department of Computer-Aided Drug Discovery, Albany Molecular Research, Inc, Bothell Research Center, 18804 North Creek Parkway, Bothell, Washington 98011, USA.
Journal of Chemical Information and Computer Sciences
|October 16, 2002
Summary
The median partitioning (MP) method effectively classifies molecules by biological activity, grouping up to 82% of active compounds into pure partitions. This computational approach aids in identifying molecules with desired activities from large databases.
Area of Science:
- Computational chemistry
- Cheminformatics
- Drug discovery
Background:
- The median partitioning (MP) method was initially designed for selecting diverse compound subsets from large databases.
- MP utilizes property descriptors to partition compounds and select representative molecules, often coupled with genetic algorithms for analysis.
- Its scalability to very large compound pools is a key advantage due to its independence from pairwise molecule comparisons.
Purpose of the Study:
- To evaluate the efficacy of the median partitioning (MP) method for classifying molecules based on their biological activity.
- To compare the performance of MP classification against traditional cell-based partitioning methods.
- To explore the utility of MP in identifying compounds with specific biological functions.
Main Methods:
- Applied the median partitioning (MP) method to classify 317 molecules across 21 distinct biological activity classes.
- Conducted MP classification calculations with and without a background set of 2000 randomly selected molecules.
- Compared the performance of MP classification against established cell-based partitioning techniques.
Main Results:
- MP classification achieved high accuracy, with up to 82% of active molecules correctly assigned to "pure" partitions sharing the same activity.
- The method demonstrated comparable or superior performance to cell-based partitioning.
- MP classification relies on direct and sequential contributions of molecular property descriptors, distinguishing it from other methods.
Conclusions:
- The median partitioning (MP) method is a robust tool not only for selecting diverse compound subsets but also for accurate biological activity classification.
- MP facilitates the identification and searching of molecules possessing desired biological activities.
- This computational approach offers a valuable strategy for accelerating drug discovery and development processes.