BICoB2014

Computational Experiment Planning and the Future of Big Data
The challenges and opportunities of Big Data are transforming the biomedical sciences. This is a long-term change still in its early days, and it makes sense to ask whether our traditional problem definitions fully capture the new possibilities. Specifically, the principal focus on data mining defines the problem as searching for models that best fit a dataset; i.e. the model is the variable, whereas the data are fixed (constant, in a given analysis). By contrast, in science the main question is "what's the best experiment to do next?" i.e. the key variable is what data to collect. Drawing from a series of genomics and computational examples, I will make the case that it is very useful to expand the problem definition of Big Data analysis in this way. Specifically, we gain major benefits from rigorously assessing the information value of the many possible "experiments" (both computational and wetlab) that we could do, in order to identify the ones that will yield the most discovery relative to their cost. Automating this assessment makes it applicable to Big Data. I will sketch the computations for doing this, and how they differ from standard data mining. I will present concrete examples of such computational experiment planning, for both a real genomics experiment design (phenotype sequencing), and for computational problems.

Chris Lee is Professor of Chemistry & Biochemistry, and Computer Science at UCLA. He has worked on a range of bioinformatics research topics since 1989, ranging from structural bioinformatics, to genomics, transcriptomics, sequence analysis, evolution, and computational experiment planning. He is founder and chair of the Bioinformatics Ph.D. program at UCLA. He received a Ph.D. in Structural Biology from Stanford (studying with Michael Levitt), and has received honors such as the MIT Technology Review TR100 award.

Analysis, Design, and Operation of Purposeful Systems: An Overview of System Relationships and Interactions