Functional Annotation of Genes through Statistical Analysis of Biomedical Articles

One of the most elaborate and important tasks in biology is the functional annotation of genes. Biologists have developed standardized and structured vocabularies, called bio-ontologies, to assist them in describing the different functions. A critical issue in the assignment of functions to genes is the utilization of knowledge from published biomedical articles. The purpose of this paper is to present a unified and comprehensive statistical methodology for functionally annotating genes using biomedical literature. Specifically, classification models are built using the discriminant analysis method while validation, analysis and interpretation of the results is based on graphical methods and various performance metrics and techniques. The general conclusions from the study are very promising, in the sense that the proposed methodology not only performs well in the assignment of functions to genes, but also provides useful and interpretable results regarding the discriminating power of certain keywords in the texts