Abstract
The classification of protein structures is essential for their function determination in bioinformatics. At present time, one can achieve high prediction accuracy easily from primary amino acid sequences. However, for further classification into various folding categories, presents a challenge to large number of folds. Recently study yielded high prediction accuracy of 65% on an independent set of 27 most populated folds. In this work, we combine data fusion scheme and a hierarchical learning architecture (HLA) and apply it on the data set gathered by Ding and Dubchak[12]. We are able to achieve an overall accuracy of 69.6%. We demonstrate that data fusion is a simple and useful scheme and could be applied to various fields.