Classification of unaligned sequences based on prototype motifs representation

Estimating the function of unknown proteins is one of the main goals in bioinformatics. In the last few years, many pattern recognition algorithms have been developed, usually focusing on global information of the protein. Conversely, predictions can be done through the identification of functional sub-sequence patterns or motifs, but most methods for motifs detection require a predefined fixed window size and aligned sequences. To confront these problems we proposed a method based on variable length motifs detection using the continuous wavelet transform and a dissimilarity space representation for evaluating the performance of our approach. A Support Vector Machine classifier is used as predictor. The highest Sn = 96.67% and Sp = 96.79%, across 10-fold cross validation, is obtained for proteins kinases. Average results Sn = 81.17% and Sp = 87.43% are comparable to current state of the art.