permGPU: Using graphics processing units in RNA microarray association studies.

BACKGROUND: Many analyses of microarray association studies involve permutation, bootstrap resampling and cross-validation, that are ideally formulated as embarrassingly parallel computing problems. Given that these analyses are computationally intensive, scalable approaches that can take advantage of multi-core processor systems need to be developed. RESULTS: We have developed a CUDA based implementation, permGPU, that employs graphics processing units in microarray association studies. We illustrate the performance and applicability of permGPU within the context of permutation resampling for a number of test statistics. An extensive simulation study demonstrates a dramatic increase in performance when using permGPU on an NVIDIA GTX 280 card compared to an optimized C/C++ solution running on a conventional Linux server. CONCLUSIONS: permGPU is available as an open-source stand-alone application and as an extension package for the R statistical environment. It provides a dramatic increase in performance for permutation resampling analysis in the context of microarray association studies. The current version offers six test statistics for carrying out permutation resampling analyses for binary, quantitative and censored time-to-event traits.

Distributed Feature Selection in Large n and Large p Regression Problems

Fitting statistical models is computationally challenging when the sample size or the dimension of the dataset is huge. An attractive approach for down-scaling the problem size is to first partition the dataset into subsets and then fit using distributed algorithms. The dataset can be partitioned either horizontally (in the sample space) or vertically (in the feature space), and the challenge arise in defining an algorithm with low communication, theoretical guarantees and excellent practical performance in general settings. For sample space partitioning, I propose a MEdian Selection Subset AGgregation Estimator ({\em message}) algorithm for solving these issues. The algorithm applies feature selection in parallel for each subset using regularized regression or Bayesian variable selection method, calculates the `median' feature inclusion index, estimates coefficients for the selected features in parallel for each subset, and then averages these estimates. The algorithm is simple, involves very minimal communication, scales efficiently in sample size, and has theoretical guarantees. I provide extensive experiments to show excellent performance in feature selection, estimation, prediction, and computation time relative to usual competitors.

While sample space partitioning is useful in handling datasets with large sample size, feature space partitioning is more effective when the data dimension is high. Existing methods for partitioning features, however, are either vulnerable to high correlations or inefficient in reducing the model dimension. In the thesis, I propose a new embarrassingly parallel framework named {\em DECO} for distributed variable selection and parameter estimation. In {\em DECO}, variables are first partitioned and allocated to m distributed workers. The decorrelated subset data within each worker are then fitted via any algorithm designed for high-dimensional problems. We show that by incorporating the decorrelation step, DECO can achieve consistent variable selection and parameter estimation on each subset with (almost) no assumptions. In addition, the convergence rate is nearly minimax optimal for both sparse and weakly sparse models and does NOT depend on the partition number m. Extensive numerical experiments are provided to illustrate the performance of the new framework.

For datasets with both large sample sizes and high dimensionality, I propose a new "divided-and-conquer" framework {\em DEME} (DECO-message) by leveraging both the {\em DECO} and the {\em message} algorithm. The new framework first partitions the dataset in the sample space into row cubes using {\em message} and then partition the feature space of the cubes using {\em DECO}. This procedure is equivalent to partitioning the original data matrix into multiple small blocks, each with a feasible size that can be stored and fitted in a computer in parallel. The results are then synthezied via the {\em DECO} and {\em message} algorithm in a reverse order to produce the final output. The whole framework is extremely scalable.

Conductive junctions with parallel graphene sheets.

The establishment of conductive graphene-molecule-graphene junction is investigated through first-principles electronic structure calculations and quantum transport calculations. The junction consists of a conjugated molecule connecting two parallel graphene sheets. The effects of molecular electronic states, structure relaxation, and molecule-graphene contact on the conductance of the junction are explored. A conductance as large as 0.38 conductance quantum is found achievable with an appropriately oriented dithiophene bridge. This work elucidates the designing principles of promising nanoelectronic devices based on conductive graphene-molecule-graphene junctions.

Understanding GPU programming for statistical computation: Studies in massively parallel massive mixtures

This article describes advances in statistical computation for large-scale data analysis in structured Bayesian mixture models via graphics processing unit (GPU) programming. The developments are partly motivated by computational challenges arising in fitting models of increasing heterogeneity to increasingly large datasets. An example context concerns common biological studies using high-throughput technologies generating many, very large datasets and requiring increasingly high-dimensional mixture models with large numbers of mixture components.We outline important strategies and processes for GPU computation in Bayesian simulation and optimization approaches, give examples of the benefits of GPU implementations in terms of processing speed and scale-up in ability to analyze large datasets, and provide a detailed, tutorial-style exposition that will benefit readers interested in developing GPU-based approaches in other statistical models. Novel, GPU-oriented approaches to modifying existing algorithms software design can lead to vast speed-up and, critically, enable statistical analyses that presently will not be performed due to compute time limitations in traditional computational environments. Supplementalmaterials are provided with all source code, example data, and details that will enable readers to implement and explore the GPU approach in this mixture modeling context. © 2010 American Statistical Association, Institute of Mathematical Statistics, and Interface Foundation of North America.

Independent evolutionary origins of landlocked alewife populations and rapid parallel evolution of phenotypic traits.

Alewife, Alosa pseudoharengus, populations occur in two discrete life-history variants, an anadromous form and a landlocked (freshwater resident) form. Landlocked populations display a consistent pattern of life-history divergence from anadromous populations, including earlier age at maturity, smaller adult body size, and reduced fecundity. In Connecticut (USA), dams constructed on coastal streams separate anadromous spawning runs from lake-resident landlocked populations. Here, we used sequence data from the mtDNA control region and allele frequency data from five microsatellite loci to ask whether coastal Connecticut landlocked alewife populations are independently evolved from anadromous populations or whether they share a common freshwater ancestor. We then used microsatellite data to estimate the timing of the divergence between anadromous and landlocked populations. Finally, we examined anadromous and landlocked populations for divergence in foraging morphology and used divergence time estimates to calculate the rate of evolution for foraging traits. Our results indicate that landlocked populations have evolved multiple times independently. Tests of population divergence and estimates of gene flow show that landlocked populations are genetically isolated, whereas anadromous populations exchange genes. These results support a 'phylogenetic raceme' model of landlocked alewife divergence, with anadromous populations forming an ancestral core from which landlocked populations independently diverged. Divergence time estimates suggest that landlocked populations diverged from a common anadromous ancestor no longer than 5000 years ago and perhaps as recently as 300 years ago, depending on the microsatellite mutation rate assumed. Examination of foraging traits reveals landlocked populations to have significantly narrower gapes and smaller gill raker spacings than anadromous populations, suggesting that they are adapted to foraging on smaller prey items. Estimates of evolutionary rates (in haldanes) indicate rapid evolution of foraging traits, possibly in response to changes in available resources.