Statistics Seminar
| Date/Time: | Monday, 28 Nov 2011 from 4:10 pm to 5:00 pm |
|---|---|
| Location: | Snedecor 3105 |
| Cost: | Free |
| URL: | www.stat.iastate.edu |
| Contact: | Jeanette La Grange |
| Phone: | 515-294-3440 |
| Channel: | College of Liberal Arts and Sciences |
| Categories: | Lectures |
| Actions: | Download iCal/vCal | Email Reminder |
Predicting un-observed outcomes of complex traits and diseases is relevant in many areas of application. For instance, accurate predictions of genetic predisposition to complex human traits and diseases could be used for personalized medicine, and the prediction of future performance is of great aid for management and breeding decisions in agriculture. Research has shown through controlled experiments and familial studies that many complex traits are highly heritable; suggesting that in principle accurate prediction of those traits should be possible from knowledge of individual's genotype.
Using high throughput technologies it is now possible to describe genomes in great detail. Incorporating the massive amounts of information generated by current genotyping methods into statistical models poses several challenges, including how to cope with the curse of dimensionality and, more importantly, how the learning algorithm can incorporate the complexity of genetic mechanisms that may involve complex interactions between genes and between genes and environmental conditions. In principle, interactions could be modeled parametrically with use of appropriate contrasts. However, this approach is only feasible in situations involving a relatively small number of genes and low-level interactions. An alternative is to deal with complexity using semi-parametric procedures such as Reproducing Kernel Hilbert Spaces (RKHS) regressions. After a brief overview of parametric methods for whole-genome regressions, we discuss the use of Bayesian RKHS for genome-enabled prediction of complex traits, including: a review of the methodology and of empirical evidence and a discussion of the strengths and limitations of the methodology from the perspective of its issue in genome-enabled prediction.