Zhi-Wei Guo , Ke Wang , Xiang Huang , Kun Li , Guojun Ouyang , Xu Yang , Jiayu Tan , Haihong Shi , Liangping Luo , Xincai Zhang , Min Zhang , Bo-Wei Han , Xiangming Zhai , Yingsong Wu , Fang Yang , Xue-Xi Yang , Jia Tang <<<

Preterm birth (PTB) occurs in around 11% of all births worldwide, resulting in significant morbidity and mortality for both mothers and offspring. Identification of pregnancies at risk of preterm birth in early pregnancy may help improve intervention and reduce its incidence. However, there exist few methods for PTB prediction developed with large sample size, high throughput screening and validation in independent cohorts. Here, we established a large scale, multi center, and case control study that included 2,590 pregnancies (2,072 full term and 518 preterm pregnancies) from three independent hospitals to develop a preterm birth classifier. We implemented whole genome sequencing on their plasma cfDNA and then their promoter profiling (read depth spanning from -1 KB to +1 KB around the transcriptional start site) was analyzed. Using three machine learning models and two feature selection algorithms, classifiers for predicting preterm delivery were developed. Among them, a classifier based on the support vector machine model and backward algorithm, named PTerm (Promoter profiling classifier for preterm prediction), exhibited the largest AUC value of 0.878 (0.852-0.904) following LOOCV cross validation. More importantly, PTerm exhibited good performance in three independent validation cohorts and achieved an overall AUC of 0.849 (0.831-0.866). Taken together, PTerm could be based on current noninvasive prenatal test (NIPT) data without changing its procedure or adding detection cost, which can be easily adapted for preclinical tests. <<<

Ja Hye Kim, Shinwon Hwang, Hyeonju Son, Dongsun Kim, Il Bin Kim, Myeong-Heui Kim, Nam Suk Sim, Dong Seok Kim, Yoo-Jin Ha, Junehawk Lee, Hoon-Chul Kang, Jeong Ho Lee, Sangwoo Kim <<<

Most somatic mutations that arise during normal development are present at low levels in single or multiple tissues depending on the developmental stage and affected organs. However, the effect of human developmental stages or mutations of different organs on the features of somatic mutations is still unclear. Here, we performed a systemic and comprehensive analysis of low-level somatic mutations using deep whole-exome sequencing (average read depth ~500×) of 498 multiple organ tissues with matched controls from 190 individuals. Our results showed that early clone-forming mutations shared between multiple organs were lower in number but showed higher allele frequencies than late clone-forming mutations [0.54 vs. 5.83 variants per individual; 6.17% vs. 1.5% variant allele frequency (VAF)] along with less nonsynonymous mutations and lower functional impacts. Additionally, early and late clone-forming mutations had unique mutational signatures that were distinct from mutations that originated from tumors. Compared with early clone-forming mutations that showed a clock-like signature across all organs or tissues studied, late clone-forming mutations showed organ, tissue, and cell-type specificity in the mutation counts, VAFs, and mutational signatures. In particular, analysis of brain somatic mutations showed a bimodal occurrence and temporal-lobe-specific signature. These findings provide new insights into the features of somatic mosaicism that are dependent on developmental stage and brain regions. <<<

Goutham Atla, Silvia Bonàs-Guarch, Mirabai Cuenca-Ardura, Anthony Beucher, Daniel J M Crouch, Javier Garcia-Hurtado, Ignasi Moran, T2DSystems Consortium, Manuel Irimia, Rashmi B Prasad, Anna L Gloyn, Lorella Marselli, Mara Suleiman, Thierry Berney, Eelco J P de Koning, Julie Kerr-Conte, Francois Pattou, John A Todd, Lorenzo Piemonti, Jorge Ferrer <<<

BACKGROUND: Non-coding genetic variants that influence gene transcription in pancreatic islets play a major role in the susceptibility to type 2 diabetes (T2D), and likely also contribute to type 1 diabetes (T1D) risk. For many loci, however, the mechanisms through which non-coding variants influence diabetes susceptibility are unknown.RESULTS: We examine splicing QTLs (sQTLs) in pancreatic islets from 399 human donors and observe that common genetic variation has a widespread influence on the splicing of genes with established roles in islet biology and diabetes. In parallel, we profile expression QTLs (eQTLs) and use transcriptome-wide association as well as genetic co-localization studies to assign islet sQTLs or eQTLs to T2D and T1D susceptibility signals, many of which lack candidate effector genes. This analysis reveals biologically plausible mechanisms, including the association of T2D with an sQTL that creates a nonsense isoform in ERO1B, a regulator of ER-stress and proinsulin biosynthesis. The expanded list of T2D risk effector genes reveals overrepresented pathways, including regulators of G-protein-mediated cAMP production. The analysis of sQTLs also reveals candidate effector genes for T1D susceptibility such as DCLRE1B, a senescence regulator, and lncRNA MEG3.CONCLUSIONS: These data expose widespread effects of common genetic variants on RNA splicing in pancreatic islets. The results support a role for splicing variation in diabetes susceptibility, and offer a new set of genetic targets with potential therapeutic benefit. <<<

Time series are often complex and rich in information but sparsely labeled and therefore challenging to model. In this paper, we propose a self-supervised framework for learning generalizable representations for non-stationary time series. Our approach, called Temporal Neighborhood Coding (TNC), takes advantage of the local smoothness of a signal's generative process to define neighborhoods in time with stationary properties. Using a debiased contrastive objective, our framework learns time series representations by ensuring that in the encoding space, the distribution of signals from within a neighborhood is distinguishable from the distribution of non-neighboring signals. Our motivation stems from the medical field, where the ability to model the dynamic nature of time series data is especially valuable for identifying, tracking, and predicting the underlying patients' latent states in settings where labeling data is practically impossible. We compare our method to recently developed unsupervised representation learning approaches and demonstrate superior performance on clustering and classification tasks for multiple datasets. <<<

Polymorphic inversions are ubiquitous in humans and they have been linked to both adaptation and disease. Following their discovery in Drosophila more than a century ago, inversions have proved to be more elusive than other structural variants. A wide variety of methods for the detection and genotyping of inversions have recently been developed: multiple techniques based on selective amplification by PCR, short- and long-read sequencing approaches, principal component analysis of small variant haplotypes, template strand sequencing, optical mapping, and various genome assembly methods. Many methods apply complex wet lab protocols or increasingly refined bioinformatic analyses. This review is an attempt to provide a practical summary and comparison of the methods that are in current use, with a focus on metrics such as the maximum size of segmental duplications at inversion breakpoints that each method can tolerate, the size range of inversions that they recover, their throughput, and whether the locations of putative inversions must be known beforehand. <<<

Long-read sequencing (LRS) has been around for more than a decade, but widespread adoption of the technology has been slow due to the perceived high error rates and high sequencing cost. This is changing due to the recent advancements to produce highly accurate sequences and the reducing costs. LRS promises significant improvement over short read sequencing in four major areas: (1) better detection of structural variation (2) better resolution of highly repetitive or nonunique regions (3) accurate long-range haplotype phasing and (4) the detection of base modifications natively from the sequencing data. Several successful applications of LRS have demonstrated its ability to resolve molecular diagnoses where short-read sequencing fails to identify a cause. However, the argument for increased diagnostic yield from LRS remains to be validated. Larger cohort studies may be required to establish the realistic boundaries of LRS's clinical utility and analytical validity, as well as the development of standards for clinical applications. We discuss the limitations of the current standard of care, and contrast with the applications and advantages of two major LRS platforms, PacBio and Oxford Nanopore, for molecular diagnostics of constitutional disorders, and present a critical argument about the potential of LRS in diagnostic settings. <<<

Cancer subtypes identification is one of the critical steps toward advancing personalized anti-cancerous therapies. Accumulation of a massive amount of multi-platform omics data measured across the same set of samples provides an opportunity to look into this deadly disease from several views simultaneously. Few integrative clustering approaches are developed to capture shared information from all the views to identify cancer subtypes. However, they have certain limitations. The challenge here is identifying the most relevant feature space from each omic view and systematically integrating them. Both the steps should lead toward a global clustering solution with biological significance. In this respect, a novel multi-omics clustering algorithm named RISynG (Recursive Integration of Synergised Graph-representations) is presented in this study. RISynG represents each omic view as two representation matrices that are Gramian and Laplacian. A parameterised combination function is defined to obtain a synergy matrix from these representation matrices. Then a recursive multi-kernel approach is applied to integrate the most relevant, shared, and complementary information captured via the respective synergy matrices. At last, clustering is applied to the integrated subspace. RISynG is benchmarked on five multi-omics cancer datasets taken from The Cancer Genome Atlas. The experimental results demonstrate RISynG's efficiency over the other approaches in this domain. <<<

We design a predictive layer for structured-output prediction (SOP) that can be plugged into any neural network guaranteeing its predictions are consistent with a set of predefined symbolic constraints. Our Semantic Probabilistic Layer (SPL) can model intricate correlations, and hard constraints, over a structured output space all while being amenable to end-to-end learning via maximum likelihood. SPLs combine exact probabilistic inference with logical reasoning in a clean and modular way, learning complex distributions and restricting their support to solutions of the constraint. As such, they can faithfully, and efficiently, model complex SOP tasks beyond the reach of alternative neuro-symbolic approaches. We empirically demonstrate that SPLs outperform these competitors in terms of accuracy on challenging SOP tasks including hierarchical multi-label classification, pathfinding and preference learning, while retaining perfect constraint satisfaction. <<<

Jinrong Peng , Donald E. Richards , Nigel M. Hartley , George P. Murphy , Katrien M. Devos , John E. Flintham , James Beales , Leslie J. Fish , Anthony J. Worland , Fatima Pelica , Duraialagaraja Sudhakar , Paul Christou , John W. Snape , Michael D. Gale , Nicholas P. Harberd <<<

World wheat grain yields increased substantially in the 1960s and 1970s because farmers rapidly adopted the new varieties and cultivation methods of the so-called ‘green revolution’1,2,3,4. The new varieties are shorter, increase grain yield at the expense of straw biomass, and are more resistant to damage by wind and rain3,4. These wheats are short because they respond abnormally to the plant growth hormone gibberellin. This reduced response to gibberellin is conferred by mutant dwarfing alleles at one of two Reduced height-1 (Rht-B1 and Rht-D1) loci4,5. Here we show that Rht-B1/Rht-D1 and maize dwarf-8 (d8)6,7 are orthologues of the Arabidopsis Gibberellin Insensitive (GAI) gene8,9. These genes encode proteins that resemble nuclear transcription factors and contain an SH2-like10 domain, indicating that phosphotyrosine may participate in gibberellin signalling. Six different orthologous dwarfing mutant alleles encode proteins that are altered in a conserved amino-terminal gibberellin signalling domain. Transgenic rice plants containing a mutant GAI allele give reduced responses to gibberellin and are dwarfed, indicating that mutant GAI orthologues could be used to increase yield in a wide range of crop species. <<<

Gene transcription is a highly regulated process in all animals. In Drosophila, two major transcriptional programs, housekeeping and developmental, have promoters with distinct regulatory compatibilities and nucleosome organization. However, it remains unclear how the differences in chromatin structure relate to the distinct regulatory properties and which chromatin remodelers are required for these programs. Using rapid degradation of core remodeler subunits in Drosophila melanogaster S2 cells, we demonstrate that developmental gene transcription requires SWI/SNF-type complexes, primarily to maintain distal enhancer accessibility. In contrast, wild-type-level housekeeping gene transcription requires the Iswi and Ino80 remodelers to maintain nucleosome positioning and phasing at promoters. These differential remodeler dependencies relate to different DNA-sequence-intrinsic nucleosome affinities, which favor a default ON state for housekeeping but a default OFF state for developmental gene transcription. Overall, our results demonstrate how different transcription-regulatory strategies are implemented by DNA sequence, chromatin structure, and remodeler activity. <<<

Probabilistic circuits (PCs) represent a probability distribution as a computational graph. Enforcing structural properties on these graphs guarantees that several inference scenarios become tractable. Among these properties, structured decomposability is a particularly appealing one: it enables the efficient and exact computations of the probability of complex logical formulas, and can be used to reason about the expected output of certain predictive models under missing data. This paper proposes Strudel, a simple, fast and accurate learning algorithm for structured-decomposable PCs. Compared to prior work for learning structured-decomposable PCs, Strudel delivers more accurate single PC models in fewer iterations, and dramatically scales learning when building ensembles of PCs. It achieves this scalability by exploiting another structural property of PCs, called determinism, and by sharing the same computational graph across mixture components. We show these advantages on standard density estimation benchmarks and challenging inference scenarios. <<<

Jeffrey K Ng, Pankaj Vats, Elyn Fritz-Waters, Stephanie Sarkar, Eleanor I Sams, Evin M Padhi, Zachary L Payne, Shawn Leonard, Marc A West, Chandler Prince, Lee Trani, Marshall Jansen, George Vacek, Mehrzad Samadi, Timothy T Harkins, Craig Pohl, Tychele N Turner <<<

Detection of de novo variants (DNVs) is critical for studies of disease-related variation and mutation rates. To accelerate DNV calling, we developed a graphics processing units-based workflow. We applied our workflow to whole-genome sequencing data from three parent-child sequenced cohorts including the Simons Simplex Collection (SSC), Simons Foundation Powering Autism Research (SPARK), and the 1000 Genomes Project (1000G) that were sequenced using DNA from blood, saliva, and lymphoblastoid cell lines (LCLs), respectively. The SSC and SPARK DNV callsets were within expectations for number of DNVs, percent at CpG sites, phasing to the paternal chromosome of origin, and average allele balance. However, the 1000G DNV callset was not within expectations and contained excessive DNVs that are likely cell line artifacts. Mutation signature analysis revealed 30% of 1000G DNV signatures matched B-cell lymphoma. Furthermore, we found variants in DNA repair genes and at Clinvar pathogenic or likely-pathogenic sites and significant excess of protein-coding DNVs in IGLL5; a gene known to be involved in B-cell lymphomas. Our study provides a new rapid DNV caller for the field and elucidates important implications of using sequencing data from LCLs for reference building and disease-related projects. <<<

Xiao Han, Xuanye Cao, Vanessa Aguiar-Pulido, Wei Yang, Menuka Karki, Paula Andrea Pimienta Ramirez, Robert M Cabrera, Ying Linda Lin, Bogdan J Wlodarczyk, Gary M Shaw, M Elizabeth Ross, Cuilian Zhang, Richard H Finnell, Yunping Lei <<<

Neural tube defects (NTDs) are congenital malformations resulting from abnormal embryonic development of the brain, spine, or spinal column. The genetic etiology of human NTDs remains poorly understood despite intensive investigation. CIC, homolog of the Capicua transcription repressor, has been reported to interact with ataxin-1 (ATXN1) and participate in the pathogenesis of spinocerebellar ataxia type 1. Our previous study demonstrated that CIC loss of function (LoF) variants contributed to the cerebral folate deficiency syndrome by downregulating folate receptor 1 (FOLR1) expression. Given the importance of folate transport in neural tube formation, we hypothesized that CIC variants could contribute to increased risk for NTDs by depressing embryonic folate concentrations. In this study, we examined CIC variants from whole-genome sequencing (WGS) data of 140 isolated spina bifida cases and identified eight missense variants of CIC gene. We tested the pathogenicity of the observed variants through multiple in vitro experiments. We determined that CIC variants decreased the FOLR1 protein level and planar cell polarity (PCP) pathway signaling in a human cell line (HeLa). In a murine cell line (NIH3T3), CIC loss of function variants downregulated PCP signaling. Taken together, this study provides evidence supporting CIC as a risk gene for human NTD. <<<

INTRODUCTION: In 2020, prior to COVID-19 vaccine rollout, the Brighton Collaboration created a priority list, endorsed by the World Health Organization, of potential adverse events relevant to COVID-19 vaccines. We adapted the Brighton Collaboration list to evaluate serious adverse events of special interest observed in mRNA COVID-19 vaccine trials.METHODS: Secondary analysis of serious adverse events reported in the placebo-controlled, phase III randomized clinical trials of Pfizer and Moderna mRNA COVID-19 vaccines in adults (NCT04368728 and NCT04470427), focusing analysis on Brighton Collaboration adverse events of special interest.RESULTS: Pfizer and Moderna mRNA COVID-19 vaccines were associated with an excess risk of serious adverse events of special interest of 10.1 and 15.1 per 10,000 vaccinated over placebo baselines of 17.6 and 42.2 (95 % CI -0.4 to 20.6 and -3.6 to 33.8), respectively. Combined, the mRNA vaccines were associated with an excess risk of serious adverse events of special interest of 12.5 per 10,000 vaccinated (95 % CI 2.1 to 22.9); risk ratio 1.43 (95 % CI 1.07 to 1.92). The Pfizer trial exhibited a 36 % higher risk of serious adverse events in the vaccine group; risk difference 18.0 per 10,000 vaccinated (95 % CI 1.2 to 34.9); risk ratio 1.36 (95 % CI 1.02 to 1.83). The Moderna trial exhibited a 6 % higher risk of serious adverse events in the vaccine group: risk difference 7.1 per 10,000 (95 % CI -23.2 to 37.4); risk ratio 1.06 (95 % CI 0.84 to 1.33). Combined, there was a 16 % higher risk of serious adverse events in mRNA vaccine recipients: risk difference 13.2 (95 % CI -3.2 to 29.6); risk ratio 1.16 (95 % CI 0.97 to 1.39).DISCUSSION: The excess risk of serious adverse events found in our study points to the need for formal harm-benefit analyses, particularly those that are stratified according to risk of serious COVID-19 outcomes. These analyses will require public release of participant level datasets. <<<

Motivation: Gene overlap occurs when two or more genes are encoded by the same nucleotides. This phenomenon is found in all taxonomic domains, but is particularly common in viruses, where it may provide a mechanism to increase the information content of compact genomes. The presence of overlapping reading frames (OvRFs) can skew estimates of selection based on the rates of non-synonymous and synonymous substitutions, since a substitution that is synonymous in one reading frame may be non-synonymous in another, and vice versa. Results: To understand the impact of OvRFs on molecular evolution, we implemented a versatile simulation model of nucleotide sequence evolution along a phylogeny with an arbitrary distribution of reading frames. We use a custom data structure to track the substitution rates at every nucleotide site, which is determined by the stationary nucleotide frequencies, transition bias, and the distribution of selection biases (dN/dS) in the respective reading frames. Availability and implementation: Our simulation model is implemented in the Python scripting language. All source code is released under the GNU General Public License (GPL) version 3, and is available at https://github.com/PoonLab/HexSE. <<<

The RNN-Transducer (RNN-T) framework for speech recognition has been growing in popularity, particularly for deployed real-time ASR systems, because it combines high accuracy with naturally streaming recognition. One of the drawbacks of RNN-T is that its loss function is relatively slow to compute, and can use a lot of memory. Excessive GPU memory usage can make it impractical to use RNN-T loss in cases where the vocabulary size is large: for example, for Chinese character-based ASR. We introduce a method for faster and more memory-efficient RNN-T loss computation. We first obtain pruning bounds for the RNN-T recursion using a simple joiner network that is linear in the encoder and decoder embeddings; we can evaluate this without using much memory. We then use those pruning bounds to evaluate the full, non-linear joiner network. <<<

This study contrasts results from different correlational methods for examining links between mother and child (N = 72 dyads) reports of early adolescent (M = 11.5 years) behavior problems and relationship negativity and support. Simple (Pearson) correlations revealed a consistent pattern of statistically significant associations, regardless of whether scores came from the same reporter or from different reporters. When correlations between behavior problems and relationship quality differed, within-reporter correlations were always greater in magnitude than between-reporter correlations. Dyadic (common fate) analyses designed for interdependent data decomposed within-reporter correlations into variance shared across reporters (dyadic correlations) and variance unique to specific reporters (individual correlations). Dyadic correlations were responsible for most associations between adolescent behavior problems and relationship negativity; after partitioning variance shared across reporters, no individual correlations emerged as statistically significant. In contrast, adolescent behavior problems were linked to relationship support via both shared variance and variance unique to maternal perceptions. Dyadic analyses provide a parsimonious alternative to multiple contrasts in instances when identical measures have been collected from multiple reporters. Findings from these analyses indicate that same-reporter variance bias should not be assumed in the absence of dyadic statistical analyses. <<<

近年来，随着全球经济的快速发展，计算机科学技术的迅猛进步和发展以及高校智慧管理对于科学技术应用的迫切需求，物联网应用、大数据管理、5G和云计算等前言技术都逐渐在高校智慧管理上得到了体现和普及。而人工智能技术作为当前在高校管理中最受青睐的高科技技术，已在教学研究管理、在线教育和学生校园生活优化等各个方面得到了深度应用并在取得显著成效。本文介绍人工智能现阶段在高校校园管理中的重要应用，以供参考。 <<<

Diffusion models have shown incredible capabilities as generative models; indeed, they power the current state-of-the-art models on text-conditioned image generation such as Imagen and DALL-E 2. In this work we review, demystify, and unify the understanding of diffusion models across both variational and score-based perspectives. We first derive Variational Diffusion Models (VDM) as a special case of a Markovian Hierarchical Variational Autoencoder, where three key assumptions enable tractable computation and scalable optimization of the ELBO. We then prove that optimizing a VDM boils down to learning a neural network to predict one of three potential objectives: the original source input from any arbitrary noisification of it, the original source noise from any arbitrarily noisified input, or the score function of a noisified input at any arbitrary noise level. We then dive deeper into what it means to learn the score function, and connect the variational perspective of a diffusion model explicitly with the Score-based Generative Modeling perspective through Tweedie's Formula. Lastly, we cover how to learn a conditional distribution using diffusion models via guidance. <<<