Dataset Information

Cross-species regulatory sequence activity prediction.

ABSTRACT: Machine learning algorithms trained to predict the regulatory activity of nucleic acid sequences have revealed principles of gene regulation and guided genetic variation analysis. While the human genome has been extensively annotated and studied, model organisms have been less explored. Model organism genomes offer both additional training sequences and unique annotations describing tissue and cell states unavailable in humans. Here, we develop a strategy to train deep convolutional neural networks simultaneously on multiple genomes and apply it to learn sequence predictors for large compendia of human and mouse data. Training on both genomes improves gene expression prediction accuracy on held out and variant sequences. We further demonstrate a novel and powerful approach to apply mouse regulatory models to analyze human genetic variants associated with molecular phenotypes and disease. Together these techniques unleash thousands of non-human epigenetic and transcriptional profiles toward more effective investigation of how gene regulation affects human disease.

SUBMITTER: Kelley DR

PROVIDER: S-EPMC7392335 | biostudies-literature | 2020 Jul

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

Cross-species regulatory sequence activity prediction.

Kelley David R DR

PLoS computational biology 20200720 7

Machine learning algorithms trained to predict the regulatory activity of nucleic acid sequences have revealed principles of gene regulation and guided genetic variation analysis. While the human genome has been extensively annotated and studied, model organisms have been less explored. Model organism genomes offer both additional training sequences and unique annotations describing tissue and cell states unavailable in humans. Here, we develop a strategy to train deep convolutional neural netwo ...[more]

PMID: 32687525

Similar Datasets

Project description:Vertebrate genomes share numerous conserved non-coding elements, many of which function as enhancer elements and are hypothesised to be under evolutionary constraint due to a need to be bound by combinations of sequence-specific transcription factors. In contrast, few such conserved elements can be detected between vertebrates and their closest invertebrate relatives. Despite this lack of sequence identity, cross-species transgenesis has identified some cases where non-coding DNA from invertebrates drives reporter gene expression in transgenic vertebrates in patterns reminiscent of the expression of vertebrate orthologues. Such instances are presumed to reflect the presence of conserved suites of binding sites in the regulatory regions of invertebrate and vertebrate orthologues, such that both regulatory elements can correctly interpret the trans-activating environment. Shuffling of binding sites has been suggested to lie behind loss of sequence conservation; however this has not been experimentally tested. Here we examine the underlying basis of enhancer activity for the Ciona intestinalis βγ-crystallin gene, which drives expression in the lens of transgenic vertebrates despite the Ciona lineage predating the evolution of the lens. We construct an interactive gene regulatory network (GRN) for vertebrate lens development, allowing network interactions to be robustly catalogued and conserved network components and features to be identified. We show that a small number of binding motifs are necessary for Ciona βγ-crystallin expression, and narrow down the likely factors that bind to these motifs. Several of these overlap with the conserved core of the vertebrate lens GRN, implicating these sites in cross species function. However when we test these motifs in a transgenic vertebrate they prove to be dispensable for reporter expression in the lens. These results show that current models depicting cross species enhancer function as dependent on conserved binding sites can be overly simplistic, with sound evolutionary inference requiring detailed dissection of underlying mechanisms.

Project description:BackgroundObtaining transcripts of homologs of closely related organisms and retrieving the reconstructed exon-intron patterns of the genes is a very important process during the analysis of the evolution of a protein family and the comparative analysis of the exon-intron structure of a certain gene from different species. Due to the ever-increasing speed of genome sequencing, the gap to genome annotation is growing. Thus, tools for the correct prediction and reconstruction of genes in related organisms become more and more important. The tool Scipio, which can also be used via the graphical interface WebScipio, performs significant hit processing of the output of the Blat program to account for sequencing errors, missing sequence, and fragmented genome assemblies. However, Scipio has so far been limited to high sequence similarity and unable to reconstruct short exons.ResultsScipio and WebScipio have fundamentally been extended to better reconstruct very short exons and intron splice sites and to be better suited for cross-species gene structure predictions. The Needleman-Wunsch algorithm has been implemented for the search for short parts of the query sequence that were not recognized by Blat. Those regions might either be short exons, divergent sequence at intron splice sites, or very divergent exons. We have shown the benefit and use of new parameters with several protein examples from completely different protein families in searches against species from several kingdoms of the eukaryotes. The performance of the new Scipio version has been tested in comparison with several similar tools.ConclusionsWith the new version of Scipio very short exons, terminal and internal, of even just one amino acid can correctly be reconstructed. Scipio is also able to correctly predict almost all genes in cross-species searches even if the ancestors of the species separated more than 100 Myr ago and if the protein sequence identity is below 80%. For our test cases Scipio outperforms all other software tested. WebScipio has been restructured and provides easy access to the genome assemblies of about 640 eukaryotic species. Scipio and WebScipio are freely accessible at http://www.webscipio.org.

Project description:IntroductionVarious sequencing based approaches are used to identify and characterize the activities of cis-regulatory elements in a genome-wide fashion. Some of these techniques rely on indirect markers such as histone modifications (ChIP-seq with histone antibodies) or chromatin accessibility (ATAC-seq, DNase-seq, FAIRE-seq), while other techniques use direct measures such as episomal assays measuring the enhancer properties of DNA sequences (STARR-seq) and direct measurement of the binding of transcription factors (ChIP-seq with transcription factor-specific antibodies). The activities of cis-regulatory elements such as enhancers, promoters, and repressors are determined by their sequence and secondary processes such as chromatin accessibility, DNA methylation, and bound histone markers.MethodsHere, machine learning models are employed to evaluate the accuracy with which cis-regulatory elements identified by various commonly used sequencing techniques can be predicted by their underlying sequence alone to distinguish between cis-regulatory activity that is reflective of sequence content versus secondary processes.Results and discussionModels trained and evaluated on D. melanogaster sequences identified through DNase-seq and STARR-seq are significantly more accurate than models trained on sequences identified by H3K4me1, H3K4me3, and H3K27ac ChIP-seq, FAIRE-seq, and ATAC-seq. These results suggest that the activity detected by DNase-seq and STARR-seq can be largely explained by underlying DNA sequence, independent of secondary processes. Experimentally, a subset of DNase-seq and H3K4me1 ChIP-seq sequences were tested for enhancer activity using luciferase assays and compared with previous tests performed on STARR-seq sequences. The experimental data indicated that STARR-seq sequences are substantially enriched for enhancer-specific activity, while the DNase-seq and H3K4me1 ChIP-seq sequences are not. Taken together, these results indicate that the DNase-seq approach identifies a broad class of regulatory elements of which enhancers are a subset and the associated data are appropriate for training models for detecting regulatory activity from sequence alone, STARR-seq data are best for training enhancer-specific sequence models, and H3K4me1 ChIP-seq data are not well suited for training and evaluating sequence-based models for cis-regulatory element prediction.

Dataset Information

Cross-species regulatory sequence activity prediction.

Publications

Cross-species regulatory sequence activity prediction.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets