Using Multiple Experts for Predicting Protein Secondary Structure

G. Armano, G. Mancosu, and A. Orro (Italy)


Machine Learning, Bioinformatics, Secondary StructurePrediction, Multiple Experts.


This paper describes a technique based on multiple experts, aimed at predicting protein secondary structures. Each expert embodies a genetic classifier (the guard) and a feedforward artificial neural network (the embedded classifier) the former being devoted to control the activation of the latter. Guards are entrusted with soft partitioning the input space according to some domain oriented metrics whose combination depends on the selection enforced by the underlying evolutionary environment. Embedded classifiers are entrusted with implementing the actual prediction task. Experiments have been performed on a subset of the PDBSelect datasets of proteins.

Important Links:

Go Back