Clustering a large number of compounds. 2. Using the Connection Machine

Area of Science:

• Computational chemistry
• Cheminformatics
• Drug discovery

Background:

• The National Cancer Institute (NCI) Repository contains approximately 230,000 compounds.
• These compounds are available for screening to identify potential drug candidates.
• Efficiently sampling these large chemical libraries is crucial for drug discovery efforts.

Purpose of the Study:

• To describe a project focused on extracting a representative sample of compounds from the NCI Repository.
• To detail the application of clustering techniques for this sampling task.
• To report on the implementation of a clustering program on a massively parallel computer.

Main Methods:

• Utilized a clustering program to analyze and sample the NCI compound repository.
• Implemented the clustering algorithm on the Connection Machine, a parallel processing computer.
• Leveraged the 16,000 processing elements of the Connection Machine for computational efficiency.

Main Results:

• The clustering approach successfully facilitated the extraction of a representative compound sample.
• The implementation on the Connection Machine transformed a complex task into a manageable process.
• This method enables routine and efficient analysis of large chemical datasets.

Conclusions:

• Massively parallel computing offers a powerful solution for analyzing large chemical libraries.
• Clustering is an effective strategy for generating representative compound subsets for screening.
• This approach enhances the efficiency of drug discovery by optimizing sample selection.