Basset: learning the regulatory code of the accessible genome with deep convolutional neural networks

By David R. Kelley, Jasper Snoek, Johannes Rinn. 2015

953 citations recorded by the source corpus.

The complex language of eukaryotic gene expression remains incompletely understood. Despite the importance suggested by many noncoding variants statistically associated with human disease, nearly all such variants have unknown mechanism. Here, we address this challenge using an approach based on a recent machine learning advance—deep convolutional neural networks (CNNs). We introduce an open source package Basset (https://github.com/davek44/Basset) to apply CNNs to learn the functional activity of DNA sequences from genomics data. We trained Basset on a compendium of accessible genomic sites mapped in 164 cell types by DNaseI-seq and demonstrate far greater predictive accuracy than previous methods. Basset predictions for the change in accessibility between variant alleles were far greater for GWAS SNPs that are likely to be causal relative to nearby SNPs in linkage disequilibrium with them. With Basset, a researcher can perform a single sequencing assay in their cell type of interest and simultaneously learn that cell’s chromatin accessibility code and annotate every mutation in the genome with its influence on present accessibility and latent potential for accessibility. Thus, Basset offers a powerful computational approach to annotate and interpret the noncoding genome.

Subfields: Molecular Biology

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

Citations in the corpus

Opportunities and obstacles for deep learning in biology and medicine (2017, methodology): “[177] inserted known protein-binding motifs into the centers of sequences and assessed the change in predicted accessibility.”
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): “, 2017], or its chromatin accessibility [Kelley et al., 2016], predicting the resistance of a bacterial strain to a drug [Drouin et al.”
RNA-protein binding motifs mining with a new hybrid deep learning based cross-domain knowledge integration approach (2016, methodology): “The pearson correlation coefficient values are calculated using the AUCs from 31 experiments for individual modalities [20] (Additional file 2).”
An equivariant Bayesian convolutional network predicts recombination hotspots and accurately resolves binding motifs (2018, background): “This is true also for modeling biology from sequence, where Deep Learning sequence models have been shown to outperform previous state of the art techniques (Zhou and Troyanskaya, 2015; Alipanahi et al., 2015; Kelley et al., 2016).”
Modern Hopfield Networks and Attention for Immune Repertoire Classification (2020, methodology): “DeepRC uses a 1D convolutional network h1 (LeCun et al., 1998; Hu et al., 2014; Kelley et al., 2016) that supplies a sequence-representation zi = h1(si), which acts as the values V = Z = (z1, .”
The Human Transcription Factors. (2018, neutral): “…controlled by TFs, and there is currently no algorithm that predicts open chromatin directly from sequence with both high sensitivity and precision: a leading model achieves 20%–35% sensitivity at a 20% false discovery rate (Kelley et al., 2016) and is most effective at identifying promoters.”
Maximum entropy methods for extracting the learned features of deep neural networks (2017, background): “, 2016) and chromatin states (Zhou and Troyanskaya, 2015; Kelley et al., 2016) from DNA sequence, greatly advancing the state-ofthe-art prediction rate in these fields.”
Deep Motif Dashboard: Visualizing and Understanding Genomic Sequences Using Deep Neural Networks (2016, background): “In recent years, there has been an explosion of deep learning models which have lead to groundbreaking results in many ﬁelds such as computer vision[13], natural language processing[25], and computational biology [2, 19, 27, 11, 14, 22].”
Deep learning in bioinformatics (2016, background): “[103] further complemented the mutation map with a line plot to show the maximum increases as well as the maximum decreases of prediction scores.”
DeepCpG: accurate prediction of single-cell DNA methylation states using deep learning (2016, methodology): “We adapted a gradient-based approach [55] to estimate mutational effects in a computationally efficient manner, thereby greatly reducing the compute cost compared to previous methods [29, 30, 32] (“Methods”).”
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].”
A guide to machine learning for biologists (2021, background): “Gene sequence Genome accessibility [14]”
Convolutional Kitchen Sinks for Transcription Factor Binding Site Prediction (2017, methodology): “We compare to DeepBind over other CNN methods [7, 9] due to its primary attention to DNA sequence speciﬁcity and ability to identify ﬁne grained (101 bp) locations of binding afﬁnity.”