Predicting effects of noncoding variants with deep learning–based sequence model

By Jian Zhou, Olga G. Troyanskaya. 2015

2,196 citations recorded by the source corpus.

Identifying functional effects of noncoding variants is a major challenge in human genetics. To predict the noncoding-variant effects de novo from sequence, we developed a deep learning–based algorithmic framework, DeepSEA (http://deepsea.princeton.edu/), that directly learns a regulatory sequence code from large-scale chromatin-profiling data, enabling prediction of chromatin effects of sequence alterations with single-nucleotide sensitivity. We further used this capability to improve prioritization of functional variants including expression quantitative trait loci (eQTLs) and disease-associated variants.

Subfields: Molecular Biology, Computational Biology & Bioinformatics

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Molecular Biology

In Computational Biology & Bioinformatics

Citations in the corpus

Big Bird: Transformers for Longer Sequences (2020, methodology): “Majority of diseases and other trait associated single-nucleotide polymorphism are correlated to non-coding genomic variations [110, 46].”
Simple and Scalable Predictive Uncertainty Estimation using Deep Ensembles (2016, background): “Deep neural networks (NNs) have achieved state-of-the-art performance on a wide variety of machine learning tasks [35] and are becoming increasingly popular in domains such as computer vision [32], speech recognition [25], natural language processing [42], and bioinformatics [2, 61].”
Learning Important Features Through Propagating Activation Differences (2017, background): ““In-silico muta-genesis” (Zhou & Troyanskaya, 2015) introduced virtual mutations at individual positions in a genomic sequence and quantiﬁed the their impact on the output.”
Enhancing Evolutionary Couplings with Deep Convolutional Neural Networks (2017, background): “Deep learning continues to develop as a powerful set of tools for solving an increasingly diverse range of problems, including many related to biological systems (LeCun et al., 2015; Zhou and Troyanskaya, 2015; Angermueller et al., 2016; Wang et al., 2017).”
WILDS: A Benchmark of in-the-Wild Distribution Shifts (2020, neutral): “…been fairly successful at deciphering functional DNA sequence patterns from learned representations and predicting the consequences of genetic perturbations in contexts in which the models are trained (Avsec et al., 2019; Zhou and Troyanskaya, 2015; Kelley et al., 2016; Jaganathan et al., 2019).”
Biological Sequence Modeling with Convolutional Kernel Networks (2017, neutral): “Even though their predictive performance was ﬁrst demonstrated for two-dimensional images, they have been recently successfully adopted for DNA sequence modeling [Alipanahi et al., 2015, Zhou and Troyanskaya, 2015].”
Likelihood Ratios for Out-of-Distribution Detection (2019, methodology): “In order to compare with the classiﬁer-based baselines, we build a classiﬁer using convolutional neural networks (CNNs), which are commonly used in both image and genomic sequence classiﬁcation problems (Alipanahi et al., 2015; Zhou & Troyanskaya, 2015; Busia et al., 2018; Ren et al., 2018b).”
Deep learning for computational biology (2016, neutral): “More recent work using convolutional neural networks (CNNs) allowed direct training on the DNA sequence, without the need to define features (Alipanahi et al , 2015; Zhou & Troyanskaya, 2015; Angermueller et al , 2016; Kelley et al , 2016).”
A Whole-Genome Analysis Framework for Effective Identification of Pathogenic Regulatory Variants in Mendelian Disease. (2016, methodology): “0, FATHMM-MKL,(54) Eigen,(11) andDeepSEA.(10) CADD and Eigen scores are extracted from the provided precomputed genome-wide file.”
Deep convolutional networks for quality assessment of protein folds (2018, background): “…learning has been applied to biological data and has yielded remarkable results for predicting the effects of genetic variations on human RNA splicing [5], for identifying DNA-and RNA-binding motifs [6], and for predicting the effects of non-coding DNA variants with single nucleotide precision [7].”
DeepChrome: deep-learning for predicting gene expression from histone modifications (2016, neutral): “, 2015), natural language processing (Collobert and Weston, 2008; Kim, 2014) and bioinformatics (Alipanahi et al., 2015; Zhou and Troyanskaya, 2015).”
RNA-protein binding motifs mining with a new hybrid deep learning based cross-domain knowledge integration approach (2016, methodology): “Here RNA sequence is one-hot representation encoded into a 101 x 4 binary matrix, whose columns correspond to A, C, G and U [8, 19].”
Sequence-based prediction of protein protein interaction using a deep-learning algorithm (2017, background): “could directly learn a regulatory sequence code from large-scale chromatin-profiling data, enabling prediction of chromatin effects of sequence alterations with single-nucleotide sensitivity [27].”
Genome-wide prediction of cis-regulatory regions using supervised deep learning methods (2016, background): “Assessing the impact of sequence variations in non-coding regions on gene expression and phenotypes is of high clinical interest [32, 61], which was one motivation for the GTEx project [62].”
SBSM-Pro: support bio-sequence machine for proteins (2023, neutral): “Classical CNNs and RNNs, as well as more recently celebrated transformer architectures like BERT [40, 41] and GPT, have their origins in domains like image analysis (for tasks such as face recognition or autonomous driving) and natural language processing.”