In PLoS One this week, researchers at Oregon Health and Science University report features of the Bayesian Network and Support Vector Machine-based machine learning techniques that are most useful for the classification of transcription factor binding sites. "Our results demonstrate good performance of classifiers both on TFBS for transcription factors used for initial training and for TFBS for other factors in cross-classification experiments," the authors write, adding that in their experience, "Bayesian network classifiers outperform SVM classifiers," overall.
Over in PLoS Biology, the University of Heidelberg's Michael Eichenlaub and Laurence Ettwiller discuss the "de novo generation of enhancers in vertebrates." Taking advantage of "the massive gene loss following the last whole genome duplication in teleosts to systematically identify regions that have lost their coding capacity but retain sequence conservation with mammals," Eichenlaub and Ettwiller found that "these regions show enhancer activity while the orthologous coding regions have no regulatory activity," suggesting to them that enhancers are an "important playground for creating new regulatory variability and evolutionary innovation," they write.
This week in PLoS Medicine, a large international collaboration led by investigators at the UK's Institute of Metabolic Science in Cambridge shows that the association of an "FTO risk allele with the odds of obesity is attenuated by 27 percent in physically active adults, highlighting the importance of PA [physical activity] in particular in those genetically predisposed to obesity." In an accompanying editorial, the University of Queensland's J. Lennert Veerman discusses the "futility of screening for genes that make you fat," saying that, among other things, the obesity-associated "rs9939609-variant of the FTO gene studied by Kilpeläinen et al. is common, but although it is the strongest known susceptibility locus for common obesity, its penetrance is low."
And in PLoS Genetics, a team led by investigators at Cornell University reports their use of a Drosophila simulans-Drosophila melanogaster genetic system, in which they identified "duplication hotspots conserved between the two species." Unlike such hotspots found in mammalian genomes, however, the Cornell-led team shows that "Drosophila duplication hotspots are not associated with sequences of high sequence identity capable of mediating non-allelic homologous recombination. Instead, Drosophila duplication hotspots are associated with late-replicating regions of the genome, suggesting a link between DNA replication and duplication rates," the authors write.