Data-mining Strategies for Personalized Medicine.pdf

1、One key issue of personalized medicine is to identify genes (proteins) responsible for drug reaction. The drug reaction generally includes the adverse drug reaction (ADR) and the therapeutic effect.
　　 In this report

2、, a data-mining methodology called chemical-protein interactome (CPI) was first proposed to identify the off-targets in adverse drug reaction (chapter 1) and the therapeutic effect (chapter 2). Through constructing and m

3、ining an in silico CPI generated by docking software, a matrix of binding strengths among a panel of drug molecules across multiple proteins was constructed. Drugs sharing same ADR outcome were found to possess similarit

4、ies in their CPI profiles towards the protein. The genetic polymorphisms within the off-targets which might enhance the unexpected interaction between drug and off-target were also identified (chapter 1). e.g., The 116th

5、 amino acid polymorphism of human HLA-B*5701 was found to enhance the unwanted binding of abacavir to the protein, which mediate the hypersensitivity of abacavir.
　　 The text-mining strategy was also applied to drill

6、 ADR-related gen es from the literature (chapter 3). An algorithm called CitationRank was introduced to prioritized genes resembling the Google PageRank algorithm, where the ADR related genes were suggested according to

7、the manner of ranking the web pages via the topology structure of the network.
　　 In conclusion, this report provides several data-mining strategies for the personalized medicine, which could enlighten the future wor