DeepSNP: An End-to-End Deep Neural Network with Attention-Based Localization for Breakpoint Detection in Single-Nucleotide Polymorphism Array Genomic Data

Clinical decision-making in cancer and other diseases relies on timely and cost-effective genome-wide testing. Classical bioinformatic algorithms, such as Rawcopy, can support genomic analysis by calling genomic breakpoints and copy-number variations (CNVs), but often require manual data curation, which is error prone, time-consuming, and thus substantially increasing costs of genomic testing and hampering timely delivery of test results to the treating physician. We aimed to investigate whether deep learning algorithms can be used to learn from genome-wide single-nucleotide polymorphism array (SNPa) data and improve state-of-the-art algorithms. We developed, applied, and validated a novel deep neural network (DNN), DeepSNP. A manually curated data set of 50 SNPa analyses was used as truth-set. We show that DeepSNP can learn from SNPa data and classify the presence or absence of genomic breakpoints within large genomic windows with high precision and recall. DeepSNP was compared with well-known neural network models as well as with Rawcopy. Moreover, the use of a localization unit indicates the ability to pinpoint genomic breakpoints despite their exact location not being provided while training. DeepSNP results demonstrate the potential of DNN architectures to learn from genomic SNPa data and encourage further adaptation for CNV detection in SNPa and other genomic data types.

Revealing turning points in ecosystem functioning over the Northern Eurasian agricultural frontier

The collapse of the Soviet Union in 1991 has been a turning point in the World history that left a unique footprint on the Northern Eurasian ecosystems. Conducting large scale mapping of environmental change and separating between naturogenic and anthropogenic drivers is a difficult endeavor in such highly complex systems. In this research a piece-wise linear regression method was used for breakpoint detection in Rain-Use Efficiency (RUE) time series and a classification of ecosystem response types was produced. Supported by earth observation data, field data, and expert knowledge, this study provides empirical evidence regarding the occurrence of drastic changes in RUE (assessment of the timing, the direction and the significance of these changes) in Northern Eurasian ecosystems between 1982 and 2011. About 36% of the study area (3.4 million km(2) ) showed significant (P < 0.05) trends and/or turning points in RUE during the observation period. A large proportion of detected turning points in RUE occurred around the fall of the Soviet Union in 1991 and in the following years which were attributed to widespread agricultural land abandonment. Our study also showed that recurrent droughts deeply affected vegetation productivity throughout the observation period, with a general worsening of the drought conditions in recent years. Moreover, recent human-induced turning points in ecosystem functioning were detected and attributed to ongoing recultivation and change in irrigation practices in the Volgograd region, and to increased salinization and increased grazing intensity around Lake Balkhash. The ecosystem-state assessment method introduced here proved to be a valuable support that highlighted hotspots of potentially altered ecosystems and allowed for disentangling human from climatic disturbances.

Virus-Clip: a fast and memory-efficient viral integration site detection tool at single-base resolution with annotation capability

Viral integration into the human genome upon infection is an important risk factor for various human malignancies. We developed viral integration site detection tool called Virus-Clip, which makes use of information extracted from soft-clipped sequencing reads to identify exact positions of human and virus breakpoints of integration events. With initial read alignment to virus reference genome and streamlined procedures, Virus-Clip delivers a simple, fast and memory-efficient solution to viral integration site detection. Moreover, it can also automatically annotate the integration events with the corresponding affected human genes. Virus-Clip has been verified using whole-transcriptome sequencing data and its detection was validated to have satisfactory sensitivity and specificity. Marked advancement in performance was detected, compared to existing tools. It is applicable to versatile types of data including whole-genome sequencing, whole-transcriptome sequencing, and targeted sequencing. Virus-Clip is available at http://web.hku.hk/~dwhho/Virus-Clip.zip.

Quantifying the effects of drought on abrupt growth decreases of major tree species in Switzerland

Drought entails important effects on tree physiology, which may result in short‐ to long‐term radial growth decreases. While the majority of studies have focused on annual drought‐related variability of growth, relatively little is known about sustained growth decreases following drought years. We apply a statistical framework to identify climatic factors that induce abrupt growth decreases and may eventually result in tree mortality. We used tree‐ring data from almost 500 standing dead trees and 200 living trees in eight sites of the Swiss network of strict forest reserves, including four of the most important Central European tree species (Abies alba, Picea abies, Fagus sylvatica and Quercus spp.). First, to assess short‐term growth responses to drought under various climate and site conditions, we calculated correlations and linear mixed‐effects models between ring‐width indices (RWIs) and drought based on the Standardized Precipitation Evapotranspiration Index (SPEI). Second, to quantify drought effects on abrupt growth decreases, we applied distributed lag nonlinear models (DLNMs), which account for both delayed effects and the nonlinear relationship between the SPEI and the occurrence of abrupt growth decreases. Positive correlations between RWIs and the SPEI indicated short‐term growth responses of all species, particularly at arid sites. Results of the DLNMs revealed species‐specific growth responses to drought. For Quercus spp., abrupt growth decreases were more likely to occur several years following severe drought, whereas for P. abies, A. alba, and F. sylvatica abrupt growth decreases started frequently immediately in the drought year. We conclude that the statistical framework allows for quantifying the effects of drought intensity on the probability of abrupt growth decreases, which ultimately contributes to an improved understanding of climate impacts on forest community dynamics.