Omics-based, Automated Disease Modeling

Jason E. Shoemaker*, Yoshihiro Kawaoka

Department of Microbiology and Immunology The Institute of Medical Science University of Tokyo

A systematic understanding of the interactions within biological systems promises to guide the design of next generation disease diagnostics and therapies. In the past decade, systems biologists have successfully applied traditional process systems and controls engineering techniques to analyze the highly complex environments of biological systems; leading to major contributions in the fields of diabetes, drug target prioritization, and metabolite optimization. Yet, many challenges remain for creating medically significant models. The most prominent of challenges is the explosion in data that has occurred since the development of microarray and next generation sequencing techniques. The wealth of data and high degree of biological variance can quickly overwhelm traditional modeling techniques.
My work currently focuses on integrating several computational techniques from the fields of systems engineering, machine learning and bioinformatics for rapidly developing disease models. Specific challenges include:
ï?· Signal deconvolution in in vivo whole genome and protein data (â??omicsâ?? data) to determine the changes local lymphocyte demographics. We have deployed a web tool, CTen, related to this work
ï?· System Inference Microarray Analysis (SIMA) in which we employ nonlinear, biologically inspired models to predict gene expression across experimental conditions
ï?· Applying machine learning to the topological characteristics biological networks to identify safe drugs
Additionally, I am involved in several systems software projects, including Garuda which aims to provide a single platform to unify systems biology tools.
The foundation of my research program will be developing machine learning and artificial intelligence algorithms that can automatically process large scale, dynamic data into effective biological models. In this poster, the successful application of the three tools mentioned above is discussed in the context of influenza virus infection and interferon therapy. Then we review methods and motives to improve these tools and how they can be implemented in long term strategy for developing omics-based, automated disease modeling platforms.