Poster: Analysis of gene ranking algorithms with extraction of relevant biomedical concepts from PubMed publications

DNA microarray is a technology that can simultaneously measure the expression levels of thousands of genes in a single experiment. Often, the most informative genes have to be selected from different gene expression level datasets. One of the possible ways to rank the genes is to use a feature selection (FS) method. There are many FS methods which can be used, but how do researches know which one is the best? Several different methods were proposed to estimate the “goodness” of the ranked gene lists [2–4]. However, these methods usually require computer experts which know how FS methods and learning algorithms work. Therefore we propose AGRA (Analysis of Gene Ranking Algorithms), a novel method where biologists and other experts with low or no previous computer knowledge can compare different FS methods with help of evidence mined from PubMed publications. To achieve this, AGRA uses the FACTA+ [1] system which is an online text search engine for MEDLINE abstracts and it helps users browse biomedical concepts (e.g. genes/proteins, diseases, symptoms, drugs, enzymes and chemical compounds) which co-occur in the documents retrieved by a search query.