Classification and analysis of regulatory pathways using graph property, biochemical and physicochemical property, and functional property

Abstract

Given a regulatory pathway system consisting of a set of proteins, can we predict which pathway class it belongs to? Such a problem is closely related to the biological function of the pathway in cells and hence is quite fundamental and essential in systems biology and proteomics. This is also an extremely difficult and challenging problem due to its complexity. To address this problem, a novel approach was developed that can be used to predict query pathways among the following six functional categories: (i) "Metabolism", (ii) "Genetic Information Processing", (iii) "Environmental Information Processing", (iv) "Cellular Processes", (v) "Organismal Systems", and (vi) "Human Diseases". The prediction method was established trough the following procedures: (i) according to the general form of pseudo amino acid composition (PseAAC), each of the pathways concerned is formulated as a 5570-D (dimensional) vector; (ii) each of components in the 5570-D vector was derived by a series of feature extractions from the pathway system according to its graphic property, biochemical and physicochemical property, as well as functional property; (iii) the minimum redundancy maximum relevance (mRMR) method was adopted to operate the prediction. A cross-validation by the jackknife test on a benchmark dataset consisting of 146 regulatory pathways indicated that an overall success rate of 78.8% was achieved by our method in identifying query pathways among the above six classes, indicating the outcome is quite promising and encouraging. To the best of our knowledge, the current study represents the first effort in attempting to identity the type of a pathway system or its biological function. It is anticipated that our report may stimulate a series of follow-up investigations in this new and challenging area.

Figure 1. The IFS curve.

The highest ACC value of IFS is 78.8% using 49 features and SMO model.

Figure 2. Distribution of the optimized 49 features.

It is straightforward to see that 25 (25/49, 51.0%) features were from the biochemical and physicochemical property and 24 (24/49, 49.0%) features were from the functional property, while none of features in graph property was selected into the optimized feature set.