10.1 Introduction: Biological Sequence Analysis and Text Mining

Today's biologists and bioinformaticians must operate effectively in a dual environment that is composed of, on the one hand, a data world, which encompasses high-throughput biological data, and, on the other hand, a knowledge world, which contains a tremendous amount of domain information encased in free text. Data mining provides a computational bridge, leading from the ever-expanding biological data world into the biological knowledge world, while text mining provides the computational means to navigate within the vast biological knowledge world currently encapsulated by the ever-growing unstructured scientific literature. Integrating these two structurally disparate worlds for biological data analysis could be advantageous. The analysis of datasets produced by large-scale biological assays would be greatly facilitated if the data mining system were cognizant of the relevant knowledge encoded in the biological literature. However, there is an apparent digital divide between data mining and text mining. The data mining methods that were developed for mining nontext biological data typically expect the source data to be highly structured, while the text mining methods have focused largely on dealing with the unstructuredness of the texts in the scientific literature, as evident from the discussions in previous chapters. In this chapter, we describe various attempts to bridge the gap by employing text as an integrated knowledge source to help guide the mining of nontext biological data. We focus on the analysis of two representative classes of nontext biological data, namely biological sequences and genetic expression profiles.