MutPred Splice: machine learning-based prediction of exonic variants that disrupt splicing

Mort, Matthew; Sterne-Weiler, Timothy; Li, Biao; Ball, Edward V.; Cooper, David Neil; Radivojac, Predrag; Sanford, Jeremy R.; Mooney, Sean D.

doi:10.1186/gb-2014-15-1-r19

Cited by 140 publications

(146 citation statements)

References 71 publications

Supporting

Mentioning

142

Contrasting

Unclassified

Order By: Relevance

“…The ability of deep learning 15 to cope with a variety of data formats has allowed handling biological sequences such as 16 DNA, RNA or amino acid directly without a need for manual feature engineering. One 17 of the challenges in bioinformatics is accurate identification of splice sites in DNA 18 sequences. The discovery of splicing has elucidated the diversity of protein production 19 and explained the increased coding potential of the genome.…”

mentioning

confidence: 99%

On the Depth of Deep Learning Models for Splice Site Identification

Elsousy

Kathiresan

Boughorbel

2018

Preprint

View full text Add to dashboard Cite

The success of deep learning has been shown in various fields including computer vision, speech recognition, natural language processing and bioinformatics. The advance of Deep Learning in Computer Vision has been an important source of inspiration for other research fields. The objective of this work is to adapt known deep learning models borrowed from computer vision such as VGGNet, Resnet and AlexNet for the classification of biological sequences. In particular, we are interested by the task of splice site identification based on raw DNA sequences. We focus on the role of model architecture depth on model training and classification performance.We show that deep learning models outperform traditional classification methods (SVM, Random Forests, and Logistic Regression) for large training sets of raw DNA sequences. Three model families are analyzed in this work namely VGGNet, AlexNet and ResNet. Three depth levels are defined for each model family. The models are benchmarked using the following metrics: Area Under ROC curve (AUC), Number of model parameters, number of floating operations. Our extensive experimental evaluation show that shallow architectures have an overall better performance than deep models. We introduced a shallow version of ResNet, named S-ResNet. We show that it gives a good trade-off between model complexity and classification performance. Author summaryDeep Learning has been widely applied to various fields in research and industry. It has 1 been also succesfully applied to genomics and in particular to splice site identification. 2We are interested in the use of advanced neural networks borrowed from computer

show abstract

mentioning

confidence: 99%

On the Depth of Deep Learning Models for Splice Site Identification

Elsousy

Kathiresan

Boughorbel

2018

Preprint

View full text Add to dashboard Cite

show abstract

“…Initial screens for relatively rare variants were carried out through PolyPhen-2 and SIFT recommended cut offs. Using Panther (http://www.pantherdb.org/pathway/) and MutPred (http://mutpred.mutdb.org/), we predicted the possible impact of a SNV on some of the 3D structural features of the protein and tried to predict potential candidates of deleterious mutations and classified them as "likely deleterious" (22,23).…”

Section: Library Preparation Formentioning

confidence: 99%

Targeted Gene Sequencing of Gallbladder Carcinoma Identifies High-impact Somatic and Rare Germline Mutations

Yadav

Sarkar

Kumari

et al. 2017

CGP

View full text Add to dashboard Cite

Abstract. Background: Gallbladder carcinoma (GBC) isGallbladder carcinoma (GBC) is a type of biliary tract cancer, characterized by rapid progression and high mortality. Significant risk factors for GBC include presence of gallstones, chronic inflammation, anomalous pancreatobiliary ductal junctions, advancing age and female gender (1).Incidence and prevalence of GBC exhibit remarkable ethnic and geographical variability. This variability has been attributed to genetic predisposition and lifestyle habits or environmental exposure. The high incidence of GBC in the Indo-Gangetic belt and its poorly characterized molecular pathology have resulted in its reputation as an enigmatic Indian carcinoma (2). Until 2010, the genomic studies of GBC were limited to scanning a few common mutations in wellknown cancer-associated genes such as tumor protein p53 (TP53) and KRAS proto-oncogene, GTPase (KRAS) (3-5). Advancements in massively parallel sequencing technology, or next-generation sequencing (NGS), have propelled the area of cancer genomics and have greatly improved the understanding of tumorigenesis and tumor evolution linked with disease progression (6, 7). Application of NGS provides a means of high-throughput identification of cancer drivers and other genes that may be clinically relevant and or actionable for precision medicine (6). During the past decade, several large-scale cancer mutational landscape studies have contributed to the understanding of the existence of clonal variability within and across tumors in various common cancer types (8). However, GBC remains one of the understudied cancer types, with only a few studies on a limited number of samples (9-11). These studies and previous candidate gene-sequencing investigations indicate that GBC tumors from patients from different geographic regions may have different genetic architectures, suggesting the need for mutation exploration using a high-throughput NGS approach from varied populations (5, 12).The current treatment for patients with GBC is operative resection. However, resection is limited to cases that present at an early stage. GBCs which have progressed are largely incurable and the treatment mostly relies on focal radiation therapy and/or generic chemotherapy (12). The overall survival rate for patients with advanced GBC (stage 3 and 4) is very poor, and fewer than 10% survive for 5 years post 495 This article is freely accessible online.

show abstract

“…Les variants synonymes (des mutations ne modifiant pas l'acide aminé) étaient ainsi écartés, au même titre que les variants localisés en dehors des régions codantes. Il est maintenant établi qu'entre 15 % et 35 % des mutations connues pour être la cause d'une maladie génétique affecteraient l'épissage [11,12]. Sans a priori sur le fait qu'un variant génétique puisse être ou non une mutation causale d'une maladie génétique, certaines études estiment même à plus de 60 % la proportion de variants qui pourraient affecter l'épissage [13].…”

Section: Dérégulation De L'épissage Dans Les Maladies Raresunclassified

Altérations de l’épissage et maladies rares

Grange

2016

Med Sci (Paris)

View full text Add to dashboard Cite

> La majorité des gènes codant des protéines chez l'homme sont soumis à l'épissage alternatif, le principal mécanisme permettant d'augmenter la diversité du transcriptome et du protéome. La régulation de l'épissage dépend de complexes ribonucléoprotéiques qui composent le splicéosome, ainsi que de protéines régulatrices qui reconnaissent des séquences au niveau des ARN pré-messagers. De plus en plus de mutations au niveau de ces séquences, mais également au niveau de composants du splicéosome ou de facteurs d'épissage, sont décrites pour être responsables du développement de maladies rares. Depuis ces dernières années, des approches de thérapie ciblée sont développées pour restaurer au moins partiellement des événements d'épissage altérés dans le cadre de ces maladies. < La Figure 1 présente les grandes étapes de la régulation de l'expression des gènes. Un gène peut être à l'origine de la synthèse de plusieurs transcrits distincts traduits en protéines différentes, qui peuvent avoir des fonctions ou des localisations cellulaires différentes. Certains transcrits ont également un rôle régulateur et sont dégradés par la voie du NMD (nonsense-mediated mRNA decay) [2]. Ces transcrits sont reconnus par le NMD parce qu'ils présentent un codon stop pré-maturé (ou PTC pour premature stop codon). L'épissage alternatif n'est pas une exception mais plutôt une règle puisque ce mécanisme concerne la très large majorité des gènes humains codants. Si on considère les gènes avec plusieurs exons, 90 % des gènes humains codants sont soumis à épissage alternatif [3,4]. Il y a en moyenne environ quatre transcrits différents par gène [3,4] et pour un tiers des gènes, l'épissage alternatif est à l'origine de la synthèse d'au moins cinq ARN messagers différents [3,4]. L'estimation de cette diversité des ARN messagers se fonde sur les connaissances actuelles, et notamment sur les bases de données de transcrits qui restent limitées : en fonction des types cellulaires et tissulaires, des stades de développement, des pathologies et des traitements appliqués (par exemple, siARN [small interfering RNA], composés chimiques, rayonnements ultra-violet, stress cytotoxique, etc.), à peine plus de la moitié seulement des transcrits seraient mis en évi-

show abstract

MutPred Splice: machine learning-based prediction of exonic variants that disrupt splicing

Cited by 140 publications

References 71 publications

On the Depth of Deep Learning Models for Splice Site Identification

On the Depth of Deep Learning Models for Splice Site Identification

Targeted Gene Sequencing of Gallbladder Carcinoma Identifies High-impact Somatic and Rare Germline Mutations

Altérations de l’épissage et maladies rares

Contact Info

Product

Resources

About