Deep learning models predict regulatory variants in pancreatic islets and refine type 2 diabetes association signals

Wesolowska-Andersen, Agata; Yu, Grace; Nylander, Vibe; Abaitua, Fernando; Thurner, Matthias; Torres, Jason; Gloyn, Anna L.; McCarthy, Mark

doi:10.7554/elife.51503

Cited by 29 publications

(19 citation statements)

References 41 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…While positive examples are always the same in different sample sets, the negative examples are randomly picked out of the genome. The performance of the model in different regions of chromosome 21 can thus vary for different training sets (Wesolowska-Andersen et al, 2020). To investigate this variation, we set up 30 balanced 1 * datasets and train 30 CNNs separately.…”

Section: Investigating the Effect Of Random Selection Of The Negativementioning

confidence: 99%

Genome annotation across species using deep convolutional neural networks

Khodabandelou

Routhier

Mozziconacci

2020

PeerJ Computer Science

View full text Add to dashboard Cite

Application of deep neural network is a rapidly expanding field now reaching many disciplines including genomics. In particular, convolutional neural networks have been exploited for identifying the functional role of short genomic sequences. These approaches rely on gathering large sets of sequences with known functional role, extracting those sequences from whole-genome-annotations. These sets are then split into learning, test and validation sets in order to train the networks. While the obtained networks perform well on validation sets, they often perform poorly when applied on whole genomes in which the ratio of positive over negative examples can be very different than in the training set. We here address this issue by assessing the genome-wide performance of networks trained with sets exhibiting different ratios of positive to negative examples. As a case study, we use sequences encompassing gene starts from the RefGene database as positive examples and random genomic sequences as negative examples. We then demonstrate that models trained using data from one organism can be used to predict gene-start sites in a related species, when using training sets providing good genome-wide performance. This cross-species application of convolutional neural networks provides a new way to annotate any genome from existing high-quality annotations in a related reference species. It also provides a way to determine whether the sequence motifs recognised by chromatin-associated proteins in different species are conserved or not.

show abstract

Section: Investigating the Effect Of Random Selection Of The Negativementioning

confidence: 99%

Genome annotation across species using deep convolutional neural networks

Khodabandelou

Routhier

Mozziconacci

2020

PeerJ Computer Science

View full text Add to dashboard Cite

show abstract

“…The application of deep learning methods to characterize the regulatory potential of non-coding variants has been a subject of interest in recent years ( 6 , 21 , 22 ). Deep learning models exhibited great advantage when dealing with larger data set ( 60 ).…”

Section: Discussionmentioning

confidence: 99%

“…asthma enriched in lung, age at menarche enriched in uterus, body fat percentage enriched in muscle tissue. In most complex diseases, the association between traits and tissues is not always straightforward because in some cases, multiple tissues may be implicated in the etiology of the disease ( 6 , 59 ). In this study, we show several attractive applications for DeepFun.…”

Section: Discussionmentioning

confidence: 99%

“…causal variants) from background signals ( 3 ). In addition, >90% of the genetic variants identified from GWAS lie outside of protein-coding regions ( 4 ) and some are in gene deserts ( 5 ), implicating that they influence disease risk through transcriptional regulation mechanisms ( 6 ). However, the distinct transcription regulatory functions across different tissues and cell types have aggravated the challenge of variant prioritization and interpretation of variant effects on regulatory elements ( 7 ).…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Predicting regulatory variants using a dense epigenomic mapped CNN model elucidated the molecular basis of trait-tissue associations

Pei

Dai

et al. 2020

Nucleic Acids Research

View full text Add to dashboard Cite

Assessing the causal tissues of human complex diseases is important for the prioritization of trait-associated genetic variants. Yet, the biological underpinnings of trait-associated variants are extremely difficult to infer due to statistical noise in genome-wide association studies (GWAS), and because >90% of genetic variants from GWAS are located in non-coding regions. Here, we collected the largest human epigenomic map from ENCODE and Roadmap consortia and implemented a deep-learning-based convolutional neural network (CNN) model to predict the regulatory roles of genetic variants across a comprehensive list of epigenomic modifications. Our model, called DeepFun, was built on DNA accessibility maps, histone modification marks, and transcription factors. DeepFun can systematically assess the impact of non-coding variants in the most functional elements with tissue or cell-type specificity, even for rare variants or de novo mutations. By applying this model, we prioritized trait-associated loci for 51 publicly-available GWAS studies. We demonstrated that CNN-based analyses on dense and high-resolution epigenomic annotations can refine important GWAS associations in order to identify regulatory loci from background signals, which yield novel insights for better understanding the molecular basis of human complex disease. We anticipate our approaches will become routine in GWAS downstream analysis and non-coding variant evaluation.

show abstract

“…However, some have been identified as human validated therapeutic targets through coding variants that are causal for disease. 60,116,117 These loci, usually referred to by the name of the nearest genes, have substantially improved the estimation of T2D genetic predisposition. However, many other genes have been found to predispose to T2D in smallerscale studies in solitary specific populations.…”

Section: Genetics and Architecture Of T2dmentioning

confidence: 99%

Genetic predisposition in type 2 diabetes: A promising approach toward a personalized management of diabetes

Sirdah

Reading

2020

Clinical Genetics

View full text Add to dashboard Cite

Diabetes mellitus, also known simply as diabetes, has been described as a chronic and complex endocrine metabolic disorder that is a leading cause of death across the globe. It is considered a key public health problem worldwide and one of four important non-communicable diseases prioritized for intervention through world health campaigns by various international foundations. Among its four categories, Type 2 diabetes (T2D) is the commonest form of diabetes accounting for over 90% of worldwide cases. Unlike monogenic inherited disorders that are passed on in a simple pattern, T2D is a multifactorial disease with a complex etiology, where a mixture of genetic and environmental factors are strong candidates for the development of the clinical condition and pathology. The genetic factors are believed to be key predisposing determinants in individual susceptibility to T2D. Therefore, identifying the predisposing genetic variants could be a crucial step in T2D management as it may ameliorate the clinical condition and preclude complications. Through an understanding the unique genetic and environmental factors that influence the development of this chronic disease individuals can benefit from personalized approaches to treatment. We searched the literature published in three electronic databases: PubMed, Scopus and ISI Web of Science for the current status of T2D and its associated genetic risk variants and discus promising approaches toward a personalized management of this chronic, non-communicable disorder.

show abstract

Deep learning models predict regulatory variants in pancreatic islets and refine type 2 diabetes association signals

Cited by 29 publications

References 41 publications

Genome annotation across species using deep convolutional neural networks

Genome annotation across species using deep convolutional neural networks

Predicting regulatory variants using a dense epigenomic mapped CNN model elucidated the molecular basis of trait-tissue associations

Genetic predisposition in type 2 diabetes: A promising approach toward a personalized management of diabetes

Contact Info

Product

Resources

About