Evolutionary Drug Discovery

We have used genetic programming to predict interaction between chemicals (proto-drugs) and human cell wall enzymes. Training data being provided by both high throughput screening (HTS) and IC50 measurements at GSK.

Recent work include modelling bioavailability and gene expression networks.

Project Presentations

• Cercia workshop on Computational Intelligence in Cheminformatics, 2 March 2006
• British Computer Society 12 Feb 2004 Slides. Handout.
• Invited talk, European Workshop on Data Mining and Text Mining for Bioinformatics, Dubrovnik, Croatia, 22 September, 2003.
• Seminar, Essex University, 20 August 2003.
• BioGEC'2003 workshop, Chicago, 12 July 2003.
• Transferring Computer Science Research to Mining DNA chip Protein Expression. Talk and poster presentation at MIPNETS 25-27 June 2003.
• Seminar Colorado State University, 19 May 2003.
• Invited talk PRCVC Prague, April 2003. (Abstract).
• Invited talk Hybrid Information Systems, Santiago, 2002.
• Invited talk IBERAMIA, Seville, 2002. Slides.
• Knowledge Discovery meets Drug Discovery, Leuven, 23 Oct 2002.

Project Research Papers

• GP for Mining DNA chip data from cancer patients
• GP in Data Mining for Drug Discovery
• CEC 2003 Canberra, 8-12 Dec 2003.
• UCL Science (html)
• EvoBIO 2003 Essex, 14-15 April 2003.

• Closing report (IGR PDF).

Code

The genetic programming system GProc is controlled by parameters given by a file gp.ini. However these can be overridden via the command line.
The terminal and function sets are given by the contents of file prim.dat
Training data is given in gp.test. When working in classifier fusion mode, one or more additional files (which are specified by prim.dat) contain the output predictions (and the classifier confidence in it).
Output is directed to standard output.

Version 1.8b