How to Support Prediction of Amyloidogenic Regions - The Use of a GA-based Wrapper Feature Selections

摘要

In this paper, we address the problem of predicting the location of amyloidogenic regions in proteins. To support this process we used a genetic algorithm-based wrapper feature subset selection. The wrapper feature subset selection approach is about choosing a minimal subset of features that satisfies an evaluation criterion. We find that most of the machine learning algorithms taken from the WEKA software achieved no worse Accuracy over reduced dataset than over the non-reduced dataset. Moreover, research has confirmed the observations of other researchers, that amino-acids have highly position-dependent propensities.