The Ensembl project (https://www ensembl org) annotates genomes and disseminates genomic data for vertebrate species We create detailed and comprehensive annotation of gene structures, regulatory elements and variants, and enable comparative genomics by inferring the evolutionary history of genes and genomes Our integrated genomic data are made available in a variety of ways, including genome browsers, search interfaces, specialist tools such as the Ensembl Variant Effect Predictor, download files and programmatic interfaces Here, we present recent Ensembl developments including two new website portals Ensembl Rapid Release (http://rapid ensembl org) is designed to provide core tools and services for genomes as soon as possible and has been deployed to support large biodiversity sequencing projects Our SARS-CoV-2 genome browser (https://covid-19 ensembl org) integrates our own annotation with publicly available genomic data from numerous sources to facilitate the use of genomics in the international scientific response to the COVID-19 pandemic We also report on other updates to our annotation resources, tools and services All Ensembl data and software are freely available without restriction

Ensembl 2021

As a data-driven science, genomics largely utilizes machine learning to capture dependencies in data and derive novel biological hypotheses. However, the ability to extract new insights from the exponentially increasing volume of genomics data requires more expressive machine learning models. By effectively leveraging large data sets, deep learning has transformed fields such as computer vision and natural language processing. Now, it is becoming the method of choice for many genomics modelling tasks, including predicting the impact of genetic variation on gene regulatory mechanisms such as DNA accessibility and splicing. This Review describes different deep learning techniques and how they can be applied to extract biologically relevant information from large, complex genomic data sets.

Deep learning: new computational modelling techniques for genomics

Distribution shifts -- where the training distribution differs from the test distribution -- can substantially degrade the accuracy of machine learning (ML) systems deployed in the wild. Despite their ubiquity, these real-world distribution shifts are under-represented in the datasets widely used in the ML community today. To address this gap, we present WILDS, a curated collection of 8 benchmark datasets that reflect a diverse range of distribution shifts which naturally arise in real-world applications, such as shifts across hospitals for tumor identification; across camera traps for wildlife monitoring; and across time and location in satellite imaging and poverty mapping. On each dataset, we show that standard training results in substantially lower out-of-distribution than in-distribution performance, and that this gap remains even with models trained by existing methods for handling distribution shifts. This underscores the need for new training methods that produce models which are more robust to the types of distribution shifts that arise in practice. To facilitate method development, we provide an open-source package that automates dataset loading, contains default model architectures and hyperparameters, and standardizes evaluations. Code and leaderboards are available at this https URL.

WILDS: A Benchmark of in-the-Wild Distribution Shifts

MicroRNAs (miRNAs) act within Argonaute proteins to guide repression of messenger RNA targets. Although various approaches have provided insight into target recognition, the sparsity of miRNA-target affinity measurements has limited understanding and prediction of targeting efficacy. Here, we adapted RNA bind-n-seq to enable measurement of relative binding affinities between Argonaute-miRNA complexes and all sequences ≤12 nucleotides in length. This approach revealed noncanonical target sites specific to each miRNA, miRNA-specific differences in canonical target-site affinities, and a 100-fold impact of dinucleotides flanking each site. These data enabled construction of a biochemical model of miRNA-mediated repression, which was extended to all miRNA sequences using a convolutional neural network. This model substantially improved prediction of cellular repression, thereby providing a biochemical basis for quantitatively integrating miRNAs into gene-regulatory networks.

https://science.sciencemag.org/content/sci/early/2019/12/04/science.aav1741.full.pdf

The biochemical basis of microRNA targeting efficacy

A primary goal of human genetics is to identify DNA sequence variants that influence biomedical traits, particularly those related to the onset and progression of human disease. Over the past 25 years, progress in realizing this objective has been transformed by advances in technology, foundational genomic resources and analytical tools, and by access to vast amounts of genotype and phenotype data. Genetic discoveries have substantially improved our understanding of the mechanisms responsible for many rare and common diseases and driven development of novel preventative and therapeutic strategies. Medical innovation will increasingly focus on delivering care tailored to individual patterns of genetic predisposition.

/pdf/a-brief-history-of-human-disease-genetics-s7ecc5yuyc.pdf

A brief history of human disease genetics.

Predicting Splicing from Primary Sequence with Deep Learning.


 Introduction: Growing data established the importance of monitoring dynamic changes in circulating tumor DNA (ctDNA) to identify early signs of therapeutic responses, allowing for timely management of treatment to achieve more effective personalized therapy. Higher assay accuracy and consistency, and lower assay cost will support more clinical validation trials and benefit more cancer patients with non-invasive ctDNA NGS tests that can simultaneously map multiple genomic alterations at an affordable price. Method: The NVIGEN X - Precision Cancer Profiling test is a next generation sequencing (NGS) based circulating tumor DNA detection assay using the hybridization capture approach with customized gene panels. Our ctDNA NGS assay was developed with the use of high performance magnetic nanobeads, which enhances assay workflow at key steps including cfDNA extraction, NGS library preparation, and target enrichment. Experiments with individual plasma samples and DNA mutant fragments spiked in plasma samples were carried out to establish the assay performance such as sensitivity, specificity, consistency and data efficiency. NGS data QC metrics of the NVIGEN assay were compared with other assays in peer reviewed publications. Results: We developed a focus 32 gene panel that covers 144 kb of gene regions of clinical significance for breast cancer treatment monitoring and guidance, such as AKT1, ERBB2, PIK3CA, EGFR, ESR1, BRCA1/2, and CD274. Our results demonstrated the capability of NVIGEN X ctDNA NGS assay to detect rare copies (8 cp) of gene mutation at 0.07% MAF from DNA mutant fragments spiked into plasma samples. The NVIGEN X ctDNA NGS assays consistently presented 2-5% duplication rate, >80% on-target rate, <10% CV for key NGS data metrics, and on average required 1.36X paired reads per 1X unique coverage. Compared with the Roche Avenio assays (targeted, expanded and surveillance panels) as published in 2020 which on average required 9.36X paired reads per 1X unique coverage, the NVIGEN X -precision cancer profiling assays demonstrated 85% reduction in NGS data need to generate each unique coverage. Compared with the original Capp-seq data as published in the 2014 Nature Medicine paper which required in average 13.78 or 27.56 paired reads per unique coverage, the NVIGEN X assay demonstrated >90% reduction in NGS data need per unique coverage. Conclusion: The NVIGEN X - Precision Cancer Profiling assay provided high NGS assay performance with high sensitivity, specificity, and consistency, and significantly improved NGS data efficiency. This allows for dramatically reduced assay cost and will help support routine applications of ctDNA NGS tests to improve cancer patient treatment. Experiments of applying NVIGEN X assays for clinical research with patient samples are ongoing and will be presented.
 Citation Format: Aihua Fu, Wenwu Cui, Minh V. Ton, Kevan Wang, Weiwei Gu, Tianhong Li, Heather A. Parsons, Minetta C. Liu, George W. Sledge. Developing highly sensitive high NGS data efficient ctDNA detection assays for breast cancer surveillance [abstract]. In: Proceedings of the 2021 San Antonio Breast Cancer Symposium; 2021 Dec 7-10; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2022;82(4 Suppl):Abstract nr P2-01-15.

Wenwu Cui

Papers

Predicting Splicing from Primary Sequence with Deep Learning.

Abstract P2-01-15: Developing highly sensitive high NGS data efficient ctDNA detection assays for breast cancer surveillance