Knowledge discovery of gene functions and metabolic pathways

Abstract

In the biosphere, biological phenomena manifest as gene functions and metabolic pathways. A challenging problem is the representation, learning and reasoning about these biochemical reactions, relationships between genotypes and phenotypes, and their interplay. Building knowledge bases of gene functions and metabolic pathways often requires integrating various different kinds of knowledge into a single hierarchical framework. On one hand, the knowledge of metabolic pathways consists of kinetic simulation, graphical representation and databases. On the other hand, the complexity of gene functions includes QTL (quantitative trait locus) mappings and higher-level data mining analysis. This paper describes a hierarchical model of cognitive maps for representing signaling and metabolism knowledge as well as genotype-to-phenotype mappings. Cognitive maps are bi-directional graphs that can learn and reason quantitatively and qualitatively. This knowledge representation scheme, coupled with numerical and statistical packages, becomes a useful tool for understanding genomics and metabolism.