Connection Between Epigenome, Selective Mutability, Evolution, and Human Disease
Li, Harris et al., PLoS Genetics
Researchers at the Baylor College of Medicine and elsewhere propose a "connection between the epigenome, selective mutability, evolution, and human disease" based on the findings of their study on associations of structural mutability with germline DNA methylation and with non-allelic homologous recombination mediated by low-copy repeats. "Combined evidence from four human sperm methylome maps, human genome evolution, structural polymorphisms in the human population, and previous genomic and disease studies consistently points to a strong association of germline hypomethylation and genomic instability," the Baylor-led team writes.
Is Your R Parallel?
If you're planning on heading to the upcoming Bioconductor meeting at the Fed Hutchinson Cancer Research Center, here's a heads-up about a workshop you might want to catch.
It just so happens that Revolution Computing is conducting a tutorial on July 28th that will provide some much-needed guidance on how to give a shot in the arm to R jobs by illuminating techniques for using parallel programming to exploit multi-core-enabled workstations and commodity clusters.
For the uninitiated, Bioconductor is an open source and development software project that provides tools for genomic data analysis and is primarily based on the R statistical programming language. The workshop will show users how to re-code and parallelize loops in existing R code using freely available programming packages that be downloaded from the Comprehensive R Network Archive, an R-dedicated site with a huge list of FTP and Web servers that host up-to-date R code and documentation. The organizers are encouraging participants to bring their own BioConductor code so that hopefully by the end of the day, if they're not too fried from grappling with the difference between fine-grained, coarse-grained, and embarrassingly parallel programming, they can walk away with some speeded-up parallelized R scripts to bring back to the lab. The Bioconductor project was started in the fall of 2001 by developers at the Fred Hutchinson Cancer Research Center along with other members from various US and international institutions, but it was a 2004 Genome Biology paper that really put it on the map.
Revolution is also continuing its educational-cum-marketing efforts with another workshop directly following the Bioconductor meeting on July 31st. The vendor will hold a one-day high-performance computing training session in Washington, DC, that promises to walk users through an overview of the company's own parallel R packages as well as Rmpi, a kind of R interface for mpi Beowulf users, and the Simple Network of Workstations parallel programming interface for R. For academics, the cost is about $250 to get your learn on, and it might be worth it -- parallelizing anything is not exactly something you want to learn via the Sally Ann Struthers International Correspondence School or holed up in some lab with a textbook. Plus, they're going to cover key topics such as how to identify performance problems (the whole reason you're putting yourself through this to begin with), multithreading, multiprocess computing, batch queueing systems, using R on GPUs, and of course, dealing with tons of data. Happy tutorial taking.