In this paper an ensemble-based approach for predicting protein secondary structure is proposed. Aimed at generating and selecting a population of experts whose competence is limited to proper regions of the input space, the approach enforces heuristics that promote the selection of highly cohesive regions. Expert overlapping is also promoted, to get benefits from the combination of mul- tiple experts involved in a decision. A system has been devised and implemented according to the guidelines of the proposed approach and experiments have been performed on several benchmarking datasets. Experimental results demonstrate the validity of the approach, as the system has reached an accuracy comparable with those shown by the most acknowledged cutting-edge predictors.
USING GUARDED EXPERTS TO PERFORM PROTEIN SECONDARY STRUCTURE PREDICTION
ARMANO, GIULIANO;
2009-01-01
Abstract
In this paper an ensemble-based approach for predicting protein secondary structure is proposed. Aimed at generating and selecting a population of experts whose competence is limited to proper regions of the input space, the approach enforces heuristics that promote the selection of highly cohesive regions. Expert overlapping is also promoted, to get benefits from the combination of mul- tiple experts involved in a decision. A system has been devised and implemented according to the guidelines of the proposed approach and experiments have been performed on several benchmarking datasets. Experimental results demonstrate the validity of the approach, as the system has reached an accuracy comparable with those shown by the most acknowledged cutting-edge predictors.
I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
