Data mining: practical machine learning tools and techniques with Java implementations
Data mining: practical machine learning tools and techniques with Java implementations
Pattern Recognition Letters
| Hi-index | 0.10 |
This paper proposes an efficient conserved region detection method for multiple protein sequences. Instead of detecting conserved regions directly from the set of all participatory protein sequences, the proposed method separates the detection process as two stages. In the fist stage, a serial of principal component analysis (PCA) techniques are applied to infer the common ancestor protein from the participatory proteins based on a hypothetical evolutionary history. Then, wavelet transform is employed to derive conserved regions from the common ancestor protein in the second stage. The detected conserved regions are considered as the common conserved regions of the original protein sequences. A set of experiments indicate that the two stage strategy makes the proposed method not only prevents the residue divergence problem but also increases the detection accuracy and efficiency.