Population history of the Sardinian people inferred from whole-genome sequencing, bioRxiv, 2016-12-09

AbstractThe population of the Mediterranean island of Sardinia has made important contributions to genome-wide association studies of traits and diseases. The history of the Sardinian population has also been the focus of much research, and in recent ancient DNA (aDNA) studies, Sardinia has provided unique insight into the peopling of Europe and the spread of agriculture. In this study, we analyze whole-genome sequences of 3,514 Sardinians to address hypotheses regarding the founding of Sardinia and its relation to the peopling of Europe, including examining fine-scale substructure, population size history, and signals of admixture. We find the population of the mountainous Gennargentu region shows elevated genetic isolation with higher levels of ancestry associated with mainland Neolithic farmers and depleted ancestry associated with more recent Bronze Age Steppe migrations on the mainland. Notably, the Gennargentu region also has elevated levels of pre-Neolithic hunter-gatherer ancestry and increased affinity to Basque populations. Further, allele sharing with pre-Neolithic and Neolithic mainland populations is larger on the X chromosome compared to the autosome, providing evidence for a sex-biased demographic history in Sardinia. These results give new insight to the demography of ancestral Sardinians and help further the understanding of sharing of disease risk alleles between Sardinia and mainland populations.

biorxiv genetics 0-100-users 2016

SPRING a kinetic interface for visualizing high dimensional single-cell expression data, bioRxiv, 2016-11-30

MotivationSingle-cell gene expression profiling technologies can map the cell states in a tissue or organism. As these technologies become more common, there is a need for computational tools to explore the data they produce. In particular, existing data visualization approaches are imperfect for studying continuous gene expression topologies.ResultsForce-directed layouts of k-nearest-neighbor graphs can visualize continuous gene expression topologies in a manner that preserves high-dimensional relationships and allows manually exploration of different stable two-dimensional representations of the same data. We implemented an interactive web-tool to visualize single-cell data using force-directed graph layouts, called SPRING. SPRING reveals more detailed biological relationships than existing approaches when applied to branching gene expression trajectories from hematopoietic progenitor cells. Visualizations from SPRING are also more reproducible than those of stochastic visualization methods such as tSNE, a state-of-the-art tool.Availability<jatsext-link xmlnsxlink=httpwww.w3.org1999xlink ext-link-type=uri xlinkhref=httpskleintools.hms.harvard.edutoolsspring.html>httpskleintools.hms.harvard.edutoolsspring.html<jatsext-link>,<jatsext-link xmlnsxlink=httpwww.w3.org1999xlink ext-link-type=uri xlinkhref=httpsgithub.comAllonKleinLabSPRING>httpsgithub.comAllonKleinLabSPRING<jatsext-link>Contactcalebsw@gmail.com, allon_klein@hms.harvard.edu

biorxiv bioinformatics 0-100-users 2016

Leveraging uncertainty information from deep neural networks for disease detection, bioRxiv, 2016-10-29

AbstractDeep learning (DL) has revolutionized the field of computer vision and image processing. In medical imaging, algorithmic solutions based on DL have been shown to achieve high performance on tasks that previously required medical experts. However, DL-based solutions for disease detection have been proposed without methods to quantify and control their uncertainty in a decision. In contrast, a physician knows whether she is uncertain about a case and will consult more experienced colleagues if needed. Here we evaluate drop-out based Bayesian uncertainty measures for DL in diagnosing diabetic retinopathy (DR) from fundus images and show that it captures uncertainty better than straightforward alternatives. Furthermore, we show that uncertainty informed decision referral can improve diagnostic performance. Experiments across different networks, tasks and datasets show robust generalization. Depending on network capacity and taskdataset difficulty, we surpass 85% sensitivity and 80% specificity as recommended by the NHS when referring 0%–20% of the most uncertain decisions for further inspection. We analyse causes of uncertainty by relating intuitions from 2D visualizations to the high-dimensional image space. While uncertainty is sensitive to clinically relevant cases, sensitivity to unfamiliar data samples is task dependent, but can be rendered more robust.

biorxiv bioinformatics 0-100-users 2016

 

Created with the audiences framework by Jedidiah Carlson

Powered by Hugo