High-Order Pattern Discovery from Discrete-Valued Data
IEEE Transactions on Knowledge and Data Engineering
| Hi-index | 0.00 |
Sequence analysis often does not take the physicochemical properties into account. On the other hand, some of these properties could be useful in inferring the folding and functional attributes of the molecule when considered with the original sequence information. We evaluated here an analysis using multiple aligned sequences incorporating five physicochemical properties. In addition to site invariance information, we also consider the covariation or interdependence patterns between aligned sites using an information measure. We propose a method based on analyzing the expected mutual information between sites that is statistically significant with a confidence level. When summing the measured information along the aligned sites, we compare the pattern from the measure to the structural and active site of the molecule. In the experiments, the model enzyme molecule lysozyme is chosen. The aligned sequence data are evaluated based on the mapped physicochemical properties of the amino acid residues. Analysis between the original and the transformed sequence data incorporating the physicochemical properties are then compared, subtracted and visualized. From the comparisons, the plots show that some of the selected physicochemical properties in the analysis correlate to the locations of active sites and certain folding structure such as helices. The experiments generally support the useful role of incorporating additional physicochemical properties into sequence analysis, when significance of the statistical variations is taken into account.