Statistical Inference

The Statistical Inference Working Group identified and prioritized general classes of statistical genetics methods that will be supported by iPlant. These include General Linear Models, Mixed Models, Machine Learning, and Bayesian approaches. General Linear Models (GLMs) are being addressed first since they are most pertinent to the widest cross-section of plant biologists. The iPlant team has developed a multiple SNP forward regression version of general linear modeling and improved the performance of single SNP forward regression on graphics processing units (GPUs). In the multiple-GPU version of the code, the software will be specifically optimized to take advantage of the GPU features on the Texas Advanced Computing Center (TACC) computing cluster. Future work from the Statistical Inference group will include solutions on how to view and explore the large (2.5E+6 points) multidimensional data sets emerging from genetic association studies as well as how to make the results of such analyses more accessible to the general research community.