Search

Nowadays, with many e-commerce platforms conducting global business, e-commerce search systems are required to handle product retrieval under multilingual scenarios. Moreover, comparing with maintaining per-country specific e-commerce search systems, having a universal system across countries can further reduce the operational and computational costs, and facilitate business expansion to new countries. In this paper, we introduce a universal end-to-end multilingual retrieval system, and discuss our learnings and technical details when training and deploying the system to serve billion-scale product retrieval for e-commerce search. In particular, we propose a multilingual graph attention based retrieval network by leveraging recent advances in transformer-based multilingual language models and graph neural network architectures to capture the interactions between search queries and items in e-commerce search. Offline experiments on five countries data show that our algorithm outperforms the state-of-the-art baselines by 35% recall and 25% mAP on average. Moreover, the proposed model shows significant increase of conversion/revenue in online A/B experiments and has been deployed in production for multiple countries., Comment: Accepted by 2021 Annual Conference of the North American Chapter of the Association for Computational Linguistics (NAACL 2021)

Understanding the interaction between humans and mosquitoes is a critical area of study due to the phenomenal burdens on public health from mosquito-transmitted diseases. In this study, we conducted the first genome-wide association studies (GWAS) of self-reported mosquito bite reaction size (n = 84,724), itchiness caused by bites (n = 69,057), and perceived attractiveness to mosquitoes (n = 16,576). In total, 15 independent significant (P

Background: Despite evidence that genetic factors contribute to the duration of gestation and the risk of preterm birth, robust associations with genetic variants have not been identified. We used large data sets that included the gestational duration to determine possible genetic associations. Methods: We performed a genomewide association study in a discovery set of samples obtained from 43,568 women of European ancestry using gestational duration as a continuous trait and term or preterm (

Supplemental Digital Content is Available in the Text. Genome-wide association study on 11,891 females of European descent for self-reported dysmenorrhea pain severity identified a significant association that colocalises with the NGF locus., Dysmenorrhea is a common chronic pelvic pain syndrome affecting women of childbearing potential. Family studies suggest that genetic background influences the severity of dysmenorrhea, but genetic predisposition and molecular mechanisms underlying dysmenorrhea are not understood. In this study, we conduct the first genome-wide association study to identify genetic factors associated with dysmenorrhea pain severity. A cohort of females of European descent (n = 11,891) aged 18 to 45 years rated their average dysmenorrhea pain severity. We used a linear regression model adjusting for age and body mass index, identifying one genome-wide significant (P < 5 × 10−8) association (rs7523086, P = 4.1 × 10−14, effect size 0.1 [95% confidence interval, 0.074–0.126]). This single nucleotide polymorphism is colocalising with NGF, encoding nerve growth factor. The presence of one risk allele corresponds to a predicted 0.1-point increase in pain intensity on a 4-point ordinal pain scale. The putative effects on NGF function and/or expression remain unknown. However, genetic variation colocalises with active epigenetic marks in fat and ovary tissues, and expression levels in aorta tissue of a noncoding RNA flanking NGF correlate. Participants reporting extreme dysmenorrhea pain were more likely to report being positive for endometriosis, polycystic ovarian syndrome, depression, and other psychiatric disorders. Our results indicate that dysmenorrhea pain severity is partly genetically determined. NGF already has an established role in chronic pain disorders, and our findings suggest that NGF may be an important mediator for gynaecological/pelvic pain in the viscera.

Disrupted circadian rhythms and reduced sleep duration are associated with several human diseases, particularly obesity and type 2 diabetes, but little is known about the genetic factors influencing these heritable traits. We performed genome-wide association studies of self-reported chronotype (morning/evening person) and self-reported sleep duration in 128,266 White British individuals from the UK Biobank study. Sixteen variants were associated with chronotype (P

Quadratic inference functions (QIF) methodology is an important alternative to the generalized estimating equations (GEE) method in the longitudinal marginal model, as it offers higher estimation efficiency than the GEE when correlation structure is misspecified. The focus of this paper is on sample size determination and power calculation for QIF based on the Wald test in a marginal logistic model with covariates of treatment, time, and treatment-time interaction. We have made three contributions in this paper: (i) we derived formulas of sample size and power for QIF and compared their performance with those given by the GEE; (ii) we proposed an optimal scheme of sample size determination to overcome the difficulty of unknown true correlation matrix in the sense of minimal average risk; and (iii) we studied properties of both QIF and GEE sample size formulas in relation to the number of follow-up visits and found that the QIF gave more robust sample sizes than the GEE. Using numerical examples, we illustrated that without sacrificing statistical power, the QIF design leads to sample size saving and hence lower study cost in comparison with the GEE analysis. We conclude that the QIF analysis is appealing for longitudinal studies.

Background: Living kidney donors (LKD) allow for increased access to lifesaving organs for transplantation. There is a relative paucity of African American (AA) live kidney donors. The prevalence of medical disease in LKD candidates has not been well studied. We examined the medical limitations to living kidney donation in a large Midwestern transplant center. Methods: A total of 2519 adults (age ‡ 18 ) evaluated as potential LKD (PD) between January 1, 1996 and June 30, 2006 were prospectively followed until evaluation outcome (completed live donation, medical exclusion from live donation, non-medical exclusion from live donation). Logistic regression was used to examine the effect of age on donor exclusion, and chi-square tests were used to compare the likelihood of donor exclusions between racial and gender groups. Results: Sixty percent of PD were female (n = 1300), and 86% were Caucasian (CA) (n = 1862). Overall, 48.7% of PD who underwent evaluation became LKD. The odds of donation were 52% lower in AA compared to CA (OR 0.48 p < 0.001). Among PD excluded from dona- tion, the most common medical diagnoses were hypertension (HTN) (24.7%), inadequate creatinine clearance (10.6%) and a positive final crossmatch (10.5%). The rate of PD exclusion for obesity was twofold higher in AA compared to CA (12.8% vs. 5.8%, p < 0.001). Conclusions: Hypertension in PD is equally significant barrier to living kidney donation in AA and CA whereas obesity is a greater barrier in AA.

Motivation: Next-generation sequencing technologies have enabled the large-scale assessment of the impact of rare and low-frequency genetic variants for complex human diseases. Gene-level association tests are often performed to analyze rare variants, where multiple rare variants in a gene region are analyzed jointly. Applying gene-level association tests to analyze sequence data often requires integrating multiple heterogeneous sources of information (e.g. annotations, functional prediction scores, allele frequencies, genotypes and phenotypes) to determine the optimal analysis unit and prioritize causal variants. Given the complexity and scale of current sequence datasets and bioinformatics databases, there is a compelling need for more efficient software tools to facilitate these analyses. To answer this challenge, we developed RVTESTS, which implements a broad set of rare variant association statistics and supports the analysis of autosomal and X-linked variants for both unrelated and related individuals. RVTESTS also provides useful companion features for annotating sequence variants, integrating bioinformatics databases, performing data quality control and sample selection. We illustrate the advantages of RVTESTS in functionality and efficiency using the 1000 Genomes Project data. Availability and implementation: RVTESTS is available on Linux, MacOS and Windows. Source code and executable files can be obtained at https://github.com/zhanxw/rvtests Contact: zhanxw@gmail.com; goncalo@umich.edu; dajiang.liu@outlook.com Supplementary information: Supplementary data are available at Bioinformatics online.

Circadian rhythms are a nearly universal feature of living organisms and affect almost every biological process. Our innate preference for mornings or evenings is determined by the phase of our circadian rhythms. We conduct a genome-wide association analysis of self-reported morningness, followed by analyses of biological pathways and related phenotypes. We identify 15 significantly associated loci, including seven near established circadian genes (rs12736689 near RGS16, P=7.0 × 10−18; rs9479402 near VIP, P=3.9 × 10−11; rs55694368 near PER2, P=2.6 × 10−9; rs35833281 near HCRTR2, P=3.7 × 10−9; rs11545787 near RASD1, P=1.4 × 10−8; rs11121022 near PER3, P=2.0 × 10−8; rs9565309 near FBXL3, P=3.5 × 10−8. Circadian and phototransduction pathways are enriched in our results. Morningness is associated with insomnia and other sleep phenotypes; and is associated with body mass index and depression but we did not find evidence for a causal relationship in our Mendelian randomization analysis. Our findings reinforce current understanding of circadian biology and will guide future studies., Circadian rhythms and related behaviours vary across individuals. Here, a large genome-wide association study reveals common single nucleotide variants influencing whether an individual reports as being a ‘morning person' by identifying 15 significant loci, including 7 near known circadian genes.

Elevated low-density lipoprotein cholesterol (LDL-C) is a treatable, heritable risk factor for cardiovascular disease. Genome-wide association studies (GWASs) have identified 157 variants associated with lipid levels but are not well suited to assess the impact of rare and low-frequency variants. To determine whether rare or low-frequency coding variants are associated with LDL-C, we exome sequenced 2,005 individuals, including 554 individuals selected for extreme LDL-C (>98(th) or

Background Plasma triglyceride levels are heritable and are correlated with the risk of coronary heart disease. Sequencing of the protein-coding regions of the human genome (the exome) has the potential to identify rare mutations that have a large effect on phenotype. Methods We sequenced the protein-coding regions of 18,666 genes in each of 3734 participants of European or African ancestry in the Exome Sequencing Project. We conducted tests to determine whether rare mutations in coding sequence, individually or in aggregate within a gene, were associated with plasma triglyceride levels. For mutations associated with triglyceride levels, we subsequently evaluated their association with the risk of coronary heart disease in 110,970 persons. Results An aggregate of rare mutations in the gene encoding apolipoprotein C3 (APOC3) was associated with lower plasma triglyceride levels. Among the four mutations that drove this result, three were loss-of-function mutations: a nonsense mutation (R19X) and two splice-site mutations (IVS2+1G -> A and IVS3+1G -> T). The fourth was a missense mutation (A43T). Approximately 1 in 150 persons in the study was a heterozygous carrier of at least one of these four mutations. Triglyceride levels in the carriers were 39% lower than levels in noncarriers (P

Estimates of the ancestry of specific chromosomal regions in admixed individuals are useful for studies of human evolutionary history and for genetic association studies. Previously, this ancestry inference relied on high-quality genotypes from genome-wide association study (GWAS) arrays. These high-quality genotypes are not always available when samples are exome sequenced, and exome sequencing is the strategy of choice for many ongoing genetic studies. Here we show that off-target reads generated during exome-sequencing experiments can be combined with on-target reads to accurately estimate the ancestry of each chromosomal segment in an admixed individual. To reconstruct local ancestry, our method SEQMIX models aligned bases directly instead of relying on hard genotype calls. We evaluate the accuracy of our method through simulations and analysis of samples sequenced by the 1000 Genomes Project and the NHLBI Grand Opportunity Exome Sequencing Project. In African Americans, we show that local-ancestry estimates derived by our method are very similar to those derived with Illumina’s Omni 2.5M genotyping array and much improved in relation to estimates that use only exome genotypes and ignore off-target sequencing reads. Software implementing this method, SEQMIX, can be applied to analysis of human population history or used for genetic association studies in admixed individuals.

Background. Patients with acute kidney injury (AKI) requiring initiation of renal replacement therapy (RRT) have poor short- and long-term outcomes, including the development of dialysis dependence. Currently, little is known about what factors may predict renal recovery in this population. Methods. We conducted a single-center, retrospective analysis of 170 hospitalized adult patients with AKI attributed to acute tubular necrosis who required inpatient initiation of RRT. Data collection included patient characteristics, laboratory data, details of hospital course and degree of fluid overload at RRT initiation. The primary outcome was recovery of renal function to dialysis independence. Results. Within 1 year of RRT initiation, 35.9% (61/170) of patients reached the primary end point of renal recovery. The median (interquartile range) duration of RRT was 11 (3–33) days and 83.6% (51/61) recovered prior to hospital discharge. Recovering patients had significantly less fluid overload at the time of RRT initiation compared to non-recovering patients (3.5 versus 9.3%, P ¼ 0.004). In multivariate Cox proportional hazard regression analysis, a rise in percent fluid overload at dialysis initiation remained a significant negative predictor of renal recovery (hazard ratio 0.97, 95% confidence interval 0.95–1.00, P ¼ 0.024). Conclusions. In patients with AKI, a higher degree of fluid overload at RRT initiation predicts worse renal recovery at 1 year. Clinical trials are needed to determine whether interventions targeting fluid overload may improve patient and renal outcomes.

Macular degeneration is a common cause of blindness in the elderly. To identify rare coding variants associated with a large increase in risk of age-related macular degeneration (AMD), we sequenced 2,335 cases and 789 controls in 10 candidate loci (57 genes). To increase power, we augmented our control set with ancestry-matched exome-sequenced controls. An analysis of coding variation in 2,268 AMD cases and 2,268 ancestry-matched controls identified 2 large-effect rare variants: previously described p. Arg1210Cys encoded in the CFH gene (case frequency (f(case)) = 0.51%; control frequency (f(control)) = 0.02%; odds ratio (OR) = 23.11) and newly identified p. Lys155Gln encoded in the C3 gene (f(case) = 1.06%; f(control) = 0.39%; OR = 2.68). The variants suggest decreased inhibition of C3 by complement factor H, resulting in increased activation of the alternative complement pathway, as a key component of disease biology.

Clinical trials for preventative therapies are complex and costly endeavors focused on individuals likely to develop disease in a short time frame, randomizing them to treatment groups, and following them over time. In such trials, statistical power is governed by the rate of disease events in each group and cost is determined by randomization, treatment, and follow-up. Strategies that increase the rate of disease events by enrolling individuals with high risk of disease can significantly reduce study size, duration, and cost. Comprehensive study of common, complex diseases has resulted in a growing list of robustly associated genetic markers. Here, we evaluate the utility--in terms of trial size, duration, and cost--of enriching prevention trial samples by combining clinical information with genetic risk scores to identify individuals at greater risk of disease. We also describe a framework for utilizing genetic risk scores in these trials and evaluating the associated cost and time savings. With type 1 diabetes (T1D), type 2 diabetes (T2D), myocardial infarction (MI), and advanced age-related macular degeneration (AMD) as examples, we illustrate the potential and limitations of using genetic data for prevention trial design. We illustrate settings where incorporating genetic information could reduce trial cost or duration considerably, as well as settings where potential savings are negligible. Results are strongly dependent on the genetic architecture of the disease, but we also show that these benefits should increase as the list of robustly associated markers for each disease grows and as large samples of genotyped individuals become available.

It has been hypothesized that, in aggregate, rare variants in coding regions of genes explain a substantial fraction of the heritability of common diseases. We sequenced the exomes of 1,000 Danish cases with common forms of type 2 diabetes (including body mass index > 27.5 kg/m2 and hypertension) and 1,000 healthy controls to an average depth of 56×. Our simulations suggest that our study had the statistical power to detect at least one causal gene (a gene containing causal mutations) if the heritability of these common diseases was explained by rare variants in the coding regions of a limited number of genes. We applied a series of gene-based tests to detect such susceptibility genes. However, no gene showed a significant association with disease risk after we corrected for the number of genes analyzed. Thus, we could reject a model for the genetic architecture of type 2 diabetes where rare nonsynonymous variants clustered in a modest number of genes (fewer than 20) are responsible for the majority of disease risk.