Computational mutagenesis of E. coli lac repressor: insight into structure-function relationships and accurate prediction of mutant activity

Authors:
Majid Masso;Kahkeshan Hijazi;Nida Parvez;Iosif I. Vaisman
Affiliations:
Laboratory for Structural Bioinformatics, George Mason University, Manassas, VA;Laboratory for Structural Bioinformatics, George Mason University, Manassas, VA;Laboratory for Structural Bioinformatics, George Mason University, Manassas, VA;Laboratory for Structural Bioinformatics, George Mason University, Manassas, VA
Venue:
ISBRA'08 Proceedings of the 4th international conference on Bioinformatics research and applications
Year:
2008

Citing 4
Cited 1

C4.5: programs for machine learning

C4.5: programs for machine learning
The quickhull algorithm for convex hulls

ACM Transactions on Mathematical Software (TOMS)
Compositional Preferences in Quadruplets of Nearest Neighbor Residues in Protein Structures: Statistical Geometry Analysis

INTSYS '98 Proceedings of the IEEE International Joint Symposia on Intelligence and Systems
Data mining in bioinformatics using Weka

Bioinformatics

On the Bond Graphs in the Delaunay-Tetrahedra of the Simplicial Decomposition of Spatial Protein Structures

IWANN '09 Proceedings of the 10th International Work-Conference on Artificial Neural Networks: Part II: Distributed Computing, Artificial Intelligence, Bioinformatics, Soft Computing, and Ambient Assisted Living

Quantified Score

Hi-index	0.00

Visualization

Abstract

A computational mutagenesis methodology that utilizes a four-body,knowledge-based, statistical contact potential is applied toward quantifyingrelative changes (residual scores) to sequence-structure compatibility in E. colilac repressor due to single amino acid residue substitutions. We show that theseresidual scores correlate well with experimentally measured relative changes inprotein activity caused by the mutations. The approach also yields a measure ofenvironmental perturbation at every residue position in the protein caused bythe mutation (residual profile). Supervised learning with a decision tree algorithm,utilizing the residual profiles of over 4000 experimentally evaluated mutantsfor training, classifies the mutants based on activity with nearly 79% accuracywhile achieving 0.80 area under the receiver operating characteristic curve.A trained decision tree model is subsequently used to infer the levels of activityfor all remaining unexplored lac repressor mutants.