Improve The Linear Regression Model in Bioinformatics Using Text Mining

Linear regression is a commonly used approach in bioinformatics. One of the main challenge with applying linear regression in bioinformatics is that the number of regression weights needed to be determined is often at least one order of magnitude larger than the number of data points available for training. This sparse data problem often reduce the reliability in determining regression weights, and leads to a significant degradation in the regression error. In this paper, we presented a text mining approach that can effectively reduce the regression error given the sparse data. The main idea is to extract the profiles of keywords from the research articles that describe the properties of the genes involved in the regression model. These keyword profiles of genes are then used to construct the similarity measure of the genes, which will be used to regulate the assignment of regression weights. More specifically, genes with similar keyword profiles are likely to be assigned similar weights. One of the key challenges in exploiting text information in regression models is that many of the extracted keywords may not be relevant to the biological process, and as a result, the similarity measurement of genes may not reflect the true relationship among the genes in a particular biological process. To resolve this problem, we present a full Bayesian framework that automatically determines the importance of key words in determining the similarity of genes in their roles to a given biological process. Empirical studies with a real biological dataset show that the proposed Bayesian framework is effective in exploiting the text profiles of genes to reduce the regression errors.