One in four adults worldwide are either overweight or obese. Epidemiological studies indicate that the location and distribution of excess fat, rather than general adiposity, is most informative for predicting risk of obesity sequellae, including cardiometabolic disease and cancer. We performed a genome-wide association study meta-analysis of body fat distribution, measured by waist-to-hip ratio adjusted for BMI (WHRadjBMI), and identified 463 signals in 346 loci. Heritability and variant effects were generally stronger in women than men, and we found approximately one-third of all signals to be sexually dimorphic. The 5% of individuals carrying the most WHRadjBMI-increasing alleles were 1.62 times more likely than the bottom 5% to have a WHR above the thresholds used for metabolic syndrome. These data, made publicly available, will inform the biology of body fat distribution and its relationship with disease.

Primary open angle glaucoma (POAG) is a leading cause of blindness world-wide. To identify new susceptibility loci, we meta-analyzed GWAS results from 8 independent studies from the United States (3,853 cases and 33,480 controls) and investigated the most significant SNPs in two Australian studies (1,252 cases and 2,592 controls), 3 European studies (875 cases and 4,107 controls) and a Singaporean Chinese study (1,037 cases and 2,543 controls). A meta-analysis of top SNPs identified three novel loci: rs35934224[T] within TXNRD2 (odds ratio (OR) = 0.78, P = 4.05×10−11 encoding a mitochondrial protein required for redox homeostasis; rs7137828[T] within ATXN2 (OR = 1.17, P = 8.73×10−10), and rs2745572[A] upstream of FOXC1 (OR = 1.17, P = 1.76×10−10). Using RT-PCR and immunohistochemistry, we show TXNRD2 and ATXN2 expression in retinal ganglion cells and the optic nerve head. These results identify new pathways underlying POAG susceptibility and suggest novel targets for preventative therapies.

Individual risk of type 2 diabetes (T2D) is modified by perturbations of adipose mass, distribution and function. To investigate mechanisms responsible, we explored the molecular, cellular, and whole-body effects of T2D-associated alleles near KLF14. We show that KLF14 diabetes-risk alleles act in adipose tissue to reduce KLF14 expression, and modulate, in trans, expression of 385 genes. We demonstrate that, in human cellular studies, reduced KLF14 expression increases pre-adipocyte proliferation but disrupts lipogenesis, and, in mice, adipose-specific deletion of Klf14 partially recapitulates the human phenotype of insulin resistance, dyslipidemia and T2D. We show that KLF14 T2D risk-allele carriers shift body fat from gynoid to abdominal stores, and display a marked increase in adipocyte cell size: these effects on fat distribution, and the T2D-association, are female-specific. Metabolic risk associated with variation at this imprinted locus depends on both the sex of the subject, and of the parent from whom the risk-allele derives.

Counting objects in digital images is a process that should be replaced by machines. This tedious task is time consuming and prone to errors due to fatigue of human annotators. The goal is to have a system that takes as input an image and returns a count of the objects inside and justification for the prediction in the form of object localization. We repose a problem, originally posed by Lempitsky and Zisserman, to instead predict a count map which contains redundant counts based on the receptive field of a smaller regression network. The regression network predicts a count of the objects that exist inside this frame. By processing the image in a fully convolutional way each pixel is going to be accounted for some number of times, the number of windows which include it, which is the size of each window, (i.e., 32x32 = 1024). To recover the true count we take the average over the redundant predictions. Our contribution is redundant counting instead of predicting a density map in order to average over errors. We also propose a novel deep neural network architecture adapted from the Inception family of networks called the Count-ception network. Together our approach results in a 20% relative improvement (2.9 to 2.3 MAE) over the state of the art method by Xie, Noble, and Zisserman in 2016. Wepropose a novel construction of networks and train-arXiv:1703.08710v2 [cs.CV] 23 Jul 2017 1. Pre-process image by padding 2. Process image in a fully convolutional way 3. Combine all counts together into total count for image