A Method to Score Pathways Using Heuristic Rules

Pathway analysis is a critically important topic in high-throughput gene expression studies. Ranking pathways based on their scores is an important aspect in many existing analysis methods. However, when calculating the pathway score, how accurately the score captured by analyzing expression patterns observable from the pathway recapitulates the experimental context is an open research problem. In this thesis, we propose a new heuristic way to score pathways by considering known gene functions, gene interaction relationships and types of gene products. We take into account if the entity involved in the pathway is a ligand, a receptor, a signaling transmission gene, a transcription factor (TF) or a TF target gene in calculating the pathway score. For a computational experiment, we collected and analyzed a total of 22 gene expression data sets related to Non-Alcoholic Fatty Liver Disease (NAFLD) from NCBI GEO. The results generated by our methods more closely reveal what is expected from the experimental context than the results obtained by applying the conventional pathway scoring methods. Our heuristic method of exploiting gene product types in computing the pathway score may have paved a new way of helping biologists interpret high-throughput gene expression data over curated pathways