Inductive Logic Programming Used to Discover Topological Constraints in Protein Structures

Thispaperdescribesthe applicationof the Inductive Logic Programming0LP) program GOLEMto the discoveryof constraintsin the packingof beta-sheets in alpha/betaproteins. Theseconstraints(rules) have a role in understandingthe protein folding problem. Constraintswerelearnt for four featuresof beta-sheet packing: the windingdirection of two sequential strands, whethertwoconsecutivestrands packparallel or anti-parallel, whethertwostrandspackadjacently, andwhethera beta-strandis at anedge.Investigation of the learnt constraintsrevealedinterestingpatterns, someof whichwerepreviously known,others that werenovel. Novelfeatures includethe discovery:that the relationshipbetweenpairs of sequentialstrandsis in general one of decreasing size, and that more sequentialpairs of strands windin the direction out than the direction in. Weconclude that machine learninghas a useful place in molecularbiologyas a patterndiscoverytool.

• Type: Conference
• Year: 1994
• Where: ISMB
• Authors: Ross D. King, Dominic A. Clark, Jack Shirazi, Michael J. E. Sternberg
• Keywords: Computational Biology, Inductive Logic Programming0LP, ISMB 1994, Protein Folding Problem, Sequential Strands