Improved Algorithms for Reaction Mapping
Improved Algorithms for Reaction Mapping
John Crabtree, Dinesh Mehta, Thomas McKinnon, Anthony Dean
Colorado School of Mines
Atomic Reaction MappingAtomic Reaction Mapping
• Critical for the automated analysis of:• Enzymatic Reactions
• Biological Pathways
• Chemical Kinetic Mechanisms
• Reaction Classification• Database Consistency Checking
Problem FormulationProblem Formulation
• R1 + … + Rn <=> P1 + … + Pm
• Reaction mapping• Let v = vertex in reactant graph• Let w = vertex in product graph• Mapping: w1 = f(v1), w2 = f(v2)• Cost c(v1, v2) = 0 same bond state between w1, w2
• Cost c(v1, v2) = 1 different state between w1, w2
• Number of bonds formed/broken
• Given a valid chemical reaction, obtain a mapping of minimum cost.
Cut Successive LargestCut Successive Largest
• Complexity is O(n2) x (complexity of naming)• Guaranteed to be efficient• Can add chemical rules
Fewest Bonds FirstFewest Bonds First
• CHCO + CHCO <=> CO + CO + CHCH
• 11 bonds
• 00000000001 00000000010 etc
• Search all bit patterns in order of min cost
• Theorem: Given a valid chemical equation, FBF will produce a mathematically optimal mapping.
Experimental ResultsExperimental Results
• Gas Research Institute GRI-Mech 3.0• 325 Reactions; CSL 94% 2sec; FBF 100%
15sec
• KEGG / Ligand over 5,000 reactions
Applicable Success
FBF 100% 99%
CSL 100% 84%
Akutsu 45% 82%