Developing novel heuristic methods for integrative computational biology

点击翻译按钮获取中文摘要

Many theoretically excellent algorithms are inadequate for the high-throughput biological domains, due to the scale or complexity of the problem, or due to unrealistic assumptions. Forming intelligent hypotheses and developing computational models of biological systems without the possibility to evaluate them limits the potential to derive realistic models. We propose to improve scalability, robustness, sensitivity and specificity of pattern discovery and prediction algorithms, and integrate them to support a methodical approach to the "systems biology" analysis and visualization of high-throughput data in cancer research and intelligent decision support in medical informatics. The long term goal is to develop and then apply novel tools for the integration, analysis and interpretation of complex biomedical data with aim to identify testable hypothesis and build useful models. The short term goals include 1) developing scalable, probabilistic, network-based algorithm for comprehensive identification of effective biomarkers for early disease detection, improved diagnosis and prognosis, and treatment response prediction; 2) developing scalable network inference approaches to predict combination treatment options using drug target databases and screens combined with networks of physical and functional protein interactions; 3) developing planning approaches for drug synthesis. Using combination of heuristic algorithms and machine learning based parameter optimization will help to reduce search space. Probabilistic modeling will help to handle incomplete, contradictory and ambiguous information in an automated fashion. Ontologies will be used to support multiple viewpoints and contexts. Additional attention will be paid to ensure the tools are interactive, seamlessly integrate diverse data sources, and they have to scale to ultra-high dimensions, support multimodal and rapidly evolving representations, and handle incompleteness of domain theories. Performance evaluation will be carried out on cancer informatics applications, using multiple publicly available datasets. The results of this research will not only generate novel algorithms, but importantly, their application will lead to fathoming cancer at a molecular level, eventually improving quality and reducing cost of cancer diagnosis and treatment by identifying prognostic and predictive signatures that enable tailoring treatment to each individual patient. The research will advance computational approaches and their applicability to high-throughput systems biology applications. An important function of this proposal is the training of bioinformatics professionals for which there is still a severe deficit. We will release tools and resources for free academic use to enable even broader benefit.