Predicting the impact of non-coding variants on DNA methylation

Learning peptide properties with positive examples only

Deep learning can create accurate predictive models by exploiting existing large-scale experimental data, and guide the design of molecules. However, a major barrier is the requirement of both positive and negative examples in the classical supervised learning frameworks. Notably, most peptide databases come with missing information and low number of observations on negative examples, as such sequences are hard to obtain using high-throughput screening methods. To address this challenge, we solely exploit the limited known positive examples in a semi-supervised setting, and discover peptide sequences that are likely to map to certain antimicrobial properties via positive-unlabeled learning (PU). In particular, we use the two learning strategies of adapting base classifier and reliable negative identification to build deep learning models for inferring solubility, hemolysis, binding against SHP-2, and non-fouling activity of peptides, given their sequence. We evaluate the predictive performance of our PU learning method and show that by only using the positive data, it can achieve competitive performance when compared with the classical positive-negative (PN) classification approach, where there is access to both positive and negative examples.

Identification and quantification of N6-methyladenosine by chemical derivatization coupled with 19F NMR spectroscopy

N 6-Methyladenosine (6mA) is a well-known prokaryotic DNA modification that has been shown to play epigenetic roles in eukaryotic DNA. Accurate detection and quantification of 6mA are prerequisites for molecular understanding of the impact of 6mA modification on DNA. However, the existing methods have several problems, such as high false-positive rate, time-consuming and complex operating procedures. Chemical sensors for the selective detection of 6mA modification are rarely reported in the literature. Fluorinated phenylboronic acid combined with 19F NMR analysis is an effective method for determining DNA or RNA modification. In this study, we presented a simple and fast chemical method for labelling the 6th imino group of 6mA using a boric-acid-derived probe. Besides, the trifluoromethyl group of trifluoromethyl phenylboronic acid (2a) could detect 6mA modification through 19F NMR. Combined with this sensor system, 6mA modification could be detected well and quickly in 6 types of deoxynucleoside mixtures and DNA samples. Taken together, the method developed in the current study has potential for specific detection of 6mA in biological samples.

Proformer: a hybrid macaron transformer model predicts expression values from promoter sequences

The breakthrough high-throughput measurement of the cis-regulatory activity of millions of randomly generated promoters provides an unprecedented opportunity to systematically decode the cis-regulatory logic that determines the expression values. We developed an end-to-end transformer encoder architecture named Proformer to predict the expression values from DNA sequences. Proformer used a Macaron-like Transformer encoder architecture, where two half-step feed forward (FFN) layers were placed at the beginning and the end of each encoder block, and a separable 1D convolution layer was inserted after the first FFN layer and in front of the multi-head attention layer. The sliding k-mers from one-hot encoded sequences were mapped onto a continuous embedding, combined with the learned positional embedding and strand embedding (forward strand vs. reverse complemented strand) as the sequence input. Moreover, Proformer introduced multiple expression heads with mask filling to prevent the transformer models from collapsing when training on relatively small amount of data. We empirically determined that this design had significantly better performance than the conventional design such as using the global pooling layer as the output layer for the regression task. These analyses support the notion that Proformer provides a novel method of learning and enhances our understanding of how cis-regulatory sequences determine the expression values.

MLACNN: an attention mechanism-based CNN architecture for predicting genome-wide DNA methylation

Methylation is an important epigenetic regulation of methylation genes that plays a crucial role in regulating biological processes. While traditional methods for detecting methylation in biological experiments are constantly improving, the development of artificial intelligence has led to the emergence of deep learning and machine learning methods as a new trend. However, traditional machine learning-based methods rely heavily on manual feature extraction, and most deep learning methods for studying methylation extract fewer features due to their simple network structures. To address this, we propose a bottomneck network based on an attention mechanism and use new methods to ensure that the deep network can learn more effective features while minimizing overfitting. This approach enables the model to learn more features from nucleotide sequences and make better predictions of methylation. The model uses three coding methods to encode the original DNA sequence and then applies feature fusion based on attention mechanisms to obtain the best fusion method. Our results demonstrate that MLACNN outperforms previous methods and achieves more satisfactory performance.

DNA language models are powerful predictors of genome-wide variant effects

The expanding catalog of genome-wide association studies (GWAS) provides biological insights across a variety of species, but identifying the causal variants behind these associations remains a significant challenge. Experimental validation is both labor-intensive and costly, highlighting the need for accurate, scalable computational methods to predict the effects of genetic variants across the entire genome. Inspired by recent progress in natural language processing, unsupervised pretraining on large protein sequence databases has proven successful in extracting complex information related to proteins. These models showcase their ability to learn variant effects in coding regions using an unsupervised approach. Expanding on this idea, we here introduce the Genomic Pre-trained Network (GPN), a model designed to learn genome-wide variant effects through unsupervised pretraining on genomic DNA sequences. Our model also successfully learns gene structure and DNA motifs without any supervision. To demonstrate its utility, we train GPN on unaligned reference genomes of Arabidopsis thaliana and seven related species within the Brassicales order and evaluate its ability to predict the functional impact of genetic variants in A. thaliana by utilizing allele frequencies from the 1001 Genomes Project and a comprehensive database of GWAS. Notably, GPN outperforms predictors based on popular conservation scores such as phyloP and phastCons. Our predictions for A. thaliana can be visualized as sequence logos in the UCSC Genome Browser (https://genome.ucsc.edu/s/gbenegas/gpn-arabidopsis). We provide code (https://github.com/songlab-cal/gpn) to train GPN for any given species using its DNA sequence alone, enabling unsupervised prediction of variant effects across the entire genome.

Deep learning-empowered crop breeding: intelligent, efficient and promising

Crop breeding is one of the main approaches to increase crop yield and improve crop quality. However, the breeding process faces challenges such as complex data, difficulties in data acquisition, and low prediction accuracy, resulting in low breeding efficiency and long cycle. Deep learning-based crop breeding is a strategy that applies deep learning techniques to improve and optimize the breeding process, leading to accelerated crop improvement, enhanced breeding efficiency, and the development of higher-yielding, more adaptive, and disease-resistant varieties for agricultural production. This perspective briefly discusses the mechanisms, key applications, and impact of deep learning in crop breeding. We also highlight the current challenges associated with this topic and provide insights into its future application prospects.

O’Brien T, Perucca P, Ge Z, Anderson A. Applications for Deep Learning in Epilepsy Genetic Research

Epilepsy is a group of brain disorders characterised by an enduring predisposition to generate unprovoked seizures. Fuelled by advances in sequencing technologies and computational approaches, more than 900 genes have now been implicated in epilepsy. The development and optimisation of tools and methods for analysing the vast quantity of genomic data is a rapidly evolving area of research. Deep learning (DL) is a subset of machine learning (ML) that brings opportunity for novel investigative strategies that can be harnessed to gain new insights into the genomic risk of people with epilepsy. DL is being harnessed to address limitations in accuracy of long-read sequencing technologies, which improve on short-read methods. Tools that predict the functional consequence of genetic variation can represent breaking ground in addressing critical knowledge gaps, while methods that integrate independent but complimentary data enhance the predictive power of genetic data. We provide an overview of these DL tools and discuss how they may be applied to the analysis of genetic data for epilepsy research.

Learning Peptide Properties with Positive Examples Only

Deep learning can create accurate predictive models by exploiting existing large-scale experimental data, and guide the design of molecules. However, a major barrier is the requirement of both positive and negative examples in the classical supervised learning frameworks. Notably, most peptide databases come with missing information and low number of observations on negative examples, as such sequences are hard to obtain using high-throughput screening methods. To address this challenge, we solely exploit the limited known positive examples in a semi-supervised setting, and discover peptide sequences that are likely to map to certain antimicrobial properties via positive-unlabeled learning (PU). In particular, we use the two learning strategies of adapting base classifier and reliable negative identification to build deep learning models for inferring solubility, hemolysis, binding against SHP-2, and non-fouling activity of peptides, given their sequence. We evaluate the predictive performance of our PU learning method and show that by only using the positive data, it can achieve competitive performance when compared with the classical positive-negative (PN) classification approach, where there is access to both positive and negative examples.

On the prediction of non-CG DNA methylation using machine learning

DNA methylation can be detected and measured using sequencing instruments after sodium bisulfite conversion, but experiments can be expensive for large eukaryotic genomes. Sequencing nonuniformity and mapping biases can leave parts of the genome with low or no coverage, thus hampering the ability of obtaining DNA methylation levels for all cytosines. To address these limitations, several computational methods have been proposed that can predict DNA methylation from the DNA sequence around the cytosine or from the methylation level of nearby cytosines. However, most of these methods are entirely focused on CG methylation in humans and other mammals. In this work, we study, for the first time, the problem of predicting cytosine methylation for CG, CHG and CHH contexts on six plant species, either from the DNA primary sequence around the cytosine or from the methylation levels of neighboring cytosines. In this framework, we also study the cross-species prediction problem and the cross-context prediction problem (within the same species). Finally, we show that providing gene and repeat annotations allows existing classifiers to significantly improve their prediction accuracy. We introduce a new classifier called AMPS (annotation-based methylation prediction from sequence) that takes advantage of genomic annotations to achieve higher accuracy.

Applications of deep learning in understanding gene regulation

Gene regulation is a central topic in cell biology. Advances in omics technologies and the accumulation of omics data have provided better opportunities for gene regulation studies than ever before. For this reason deep learning, as a data-driven predictive modeling approach, has been successfully applied to this field during the past decade. In this article, we aim to give a brief yet comprehensive overview of representative deep-learning methods for gene regulation. Specifically, we discuss and compare the design principles and datasets used by each method, creating a reference for researchers who wish to replicate or improve existing methods. We also discuss the common problems of existing approaches and prospectively introduce the emerging deep-learning paradigms that will potentially alleviate them. We hope that this article will provide a rich and up-to-date resource and shed light on future research directions in this area.

DeepPHiC: predicting promoter-centered chromatin interactions using a novel deep learning approach

MOTIVATION

Promoter-centered chromatin interactions, which include promoter-enhancer (PE) and promoter-promoter (PP) interactions, are important to decipher gene regulation and disease mechanisms. The development of next-generation sequencing technologies such as promoter capture Hi-C (pcHi-C) leads to the discovery of promoter-centered chromatin interactions. However, pcHi-C experiments are expensive and thus may be unavailable for tissues/cell types of interest. In addition, these experiments may be underpowered due to insufficient sequencing depth or various artifacts, which results in a limited finding of interactions. Most existing computational methods for predicting chromatin interactions are based on in situ Hi-C and can detect chromatin interactions across the entire genome. However, they may not be optimal for predicting promoter-centered chromatin interactions.

RESULTS

We develop a supervised multi-modal deep learning model, which utilizes a comprehensive set of features such as genomic sequence, epigenetic signal, anchor distance, evolutionary features and DNA structural features to predict tissue/cell type-specific PE and PP interactions. We further extend the deep learning model in a multi-task learning and a transfer learning framework and demonstrate that the proposed approach outperforms state-of-the-art deep learning methods. Moreover, the proposed approach can achieve comparable prediction performance using predefined biologically relevant tissues/cell types compared to using all tissues/cell types in the pretraining especially for predicting PE interactions. The prediction performance can be further improved by using computationally inferred biologically relevant tissues/cell types in the pretraining, which are defined based on the common genes in the proximity of two anchors in the chromatin interactions.

AVAILABILITY AND IMPLEMENTATION

https://github.com/lichen-lab/DeepPHiC.

SUPPLEMENTARY INFORMATION

Supplementary data are available at Bioinformatics online.

Completing Single-Cell DNA Methylome Profiles via Transfer Learning Together With KL-Divergence

The high level of sparsity in methylome profiles obtained using whole-genome bisulfite sequencing in the case of low biological material amount limits its value in the study of systems in which large samples are difficult to assemble, such as mammalian preimplantation embryonic development. The recently developed computational methods for addressing the sparsity by imputing missing have their limits when the required minimum data coverage or profiles of the same tissue in other modalities are not available. In this study, we explored the use of transfer learning together with Kullback-Leibler (KL) divergence to train predictive models for completing methylome profiles with very low coverage (below 2%). Transfer learning was used to leverage less sparse profiles that are typically available for different tissues for the same species, while KL divergence was employed to maximize the usage of information carried in the input data. A deep neural network was adopted to extract both DNA sequence and local methylation patterns for imputation. Our study of training models for completing methylome profiles of bovine oocytes and early embryos demonstrates the effectiveness of transfer learning and KL divergence, with individual increase of 29.98 and 29.43%, respectively, in prediction performance and 38.70% increase when the two were used together. The drastically increased data coverage (43.80-73.6%) after imputation powers downstream analyses involving methylomes that cannot be effectively done using the very low coverage profiles (0.06-1.47%) before imputation.

DeepBarcoding: Deep Learning for Species Classification Using DNA Barcoding

DNA barcodes with short sequence fragments are used for species identification. Because of advances in sequencing technologies, DNA barcodes have gradually been emphasized. DNA sequences from different organisms are easily and rapidly acquired. Therefore, DNA sequence analysis tools play an increasingly crucial role in species identification. This study proposed deep barcoding, a deep learning framework for species classification by using DNA barcodes. Deep barcoding uses raw sequence data as the input to represent one-hot encoding as a one-dimensional image and uses a deep convolutional neural network with a fully connected deep neural network for sequence analysis. It can achieve an average accuracy of >90 percent for both simulation and real datasets. Although deep learning yields outstanding performance for species classification with DNA sequences, its application remains a challenge. The deep barcoding model can be a potential tool for species classification and can elucidate DNA barcode-based species identification.

Genomics enters the deep learning era

The tremendous amount of biological sequence data available, combined with the recent methodological breakthrough in deep learning in domains such as computer vision or natural language processing, is leading today to the transformation of bioinformatics through the emergence of deep genomics, the application of deep learning to genomic sequences. We review here the new applications that the use of deep learning enables in the field, focusing on three aspects: the functional annotation of genomes, the sequence determinants of the genome functions and the possibility to write synthetic genomic sequences.

Towards the Genetic Architecture of Complex Gene Expression Traits: Challenges and Prospects for eQTL Mapping in Humans

The discovery of expression quantitative trait loci (eQTLs) and their target genes (eGenes) has not only compensated for the limitations of genome-wide association studies for complex phenotypes but has also provided a basis for predicting gene expression. Efforts have been made to develop analytical methods in statistical genetics, a key discipline in eQTL analysis. In particular, mixed model– and deep learning–based analytical methods have been extremely beneficial in mapping eQTLs and predicting gene expression. Nevertheless, we still face many challenges associated with eQTL discovery. Here, we discuss two key aspects of these challenges: 1, the complexity of eTraits with various factors such as polygenicity and epistasis and 2, the voluminous work required for various types of eQTL profiles. The properties and prospects of statistical methods, including the mixed model method, Bayesian inference, the deep learning method, and the integration method, are presented as future directions for eQTL discovery. This review will help expedite the design and use of efficient methods for eQTL discovery and eTrait prediction.

Prediction of disease-associated nsSNPs by integrating multi-scale ResNet models with deep feature fusion

More than 6000 human diseases have been recorded to be caused by non-synonymous single nucleotide polymorphisms (nsSNPs). Rapid and accurate prediction of pathogenic nsSNPs can improve our understanding of the principle and design of new drugs, which remains an unresolved challenge. In the present work, a new computational approach, termed MSRes-MutP, is proposed based on ResNet blocks with multi-scale kernel size to predict disease-associated nsSNPs. By feeding the serial concatenation of the extracted four types of features, the performance of MSRes-MutP does not obviously improve. To address this, a second model FFMSRes-MutP is developed, which utilizes deep feature fusion strategy and multi-scale 2D-ResNet and 1D-ResNet blocks to extract relevant two-dimensional features and physicochemical properties. FFMSRes-MutP with the concatenated features achieves a better performance than that with individual features. The performance of FFMSRes-MutP is benchmarked on five different datasets. It achieves the Matthew's correlation coefficient (MCC) of 0.593 and 0.618 on the PredictSNP and MMP datasets, which are 0.101 and 0.210 higher than that of the existing best method PredictSNP1. When tested on the HumDiv and HumVar datasets, it achieves MCC of 0.9605 and 0.9507, and area under curve (AUC) of 0.9796 and 0.9748, which are 0.1747 and 0.2669, 0.0853 and 0.1335, respectively, higher than the existing best methods PolyPhen-2 and FATHMM (weighted). In addition, on blind test using a third-party dataset, FFMSRes-MutP performs as the second-best predictor (with MCC and AUC of 0.5215 and 0.7633, respectively), when compared with the other four predictors. Extensive benchmarking experiments demonstrate that FFMSRes-MutP achieves effective feature fusion and can be explored as a useful approach for predicting disease-associated nsSNPs. The webserver is freely available at http://csbio.njust.edu.cn/bioinf/ffmsresmutp/ for academic use.

Methylation quantitative trait locus rs5326 is associated with susceptibility and effective dosage of methadone maintenance treatment for heroin use disorder

RATIONALE

Opioid use disorder is a complicated brain disease with high heritability. The underlying mechanisms of the genetic underpinnings in the susceptibility and treatment response of opioid use disorder remain elusive.

OBJECTIVES

To reveal the potential associations of genotypes and gene methylations of dopaminergic system genes, as well as roles of them in opioid use disorder. In the present study, we detected the DNA methylation in the promoter regions of five representative dopaminergic system genes (DRD1, DRD2, SLC6A3, TH, and COMT) between 120 patients with heroin use disorder in methadone maintenance treatment (MMT) program and 111 healthy controls. The associations of 25 SNPs in the above genes and methylation of 237 CpG sites, known as methylation quantitative trait loci (mQTLs), were determined. Then, the correlations of the above mQTLs and traits of heroin use disorder were analyzed in a sample set of 801 patients with heroin use disorder and 930 healthy controls.

RESULTS

Our results demonstrated that several mQTLs in the DRD1 and DRD2 genes were identified both in the heroin use disorder and healthy control groups. Interestingly, rs4867798-CpG_174872884 and rs5326-CpG_174872884 in the DRD1 gene were the unique SNP-CpG pairs in the patients with heroin use disorder. Furthermore, mQTL rs5326 was associated with the susceptibility and effective dosage of MMT for heroin use disorder, and demonstrated allele-specific correlation with the expression of the DRD1 gene in the human caudate.

CONCLUSIONS

Our findings suggest that some mQTLs may be associated with traits of opioid use disorder by implicating the DNA methylation and gene expression.

Machine learning meets omics: applications and perspectives

The innovation of biotechnologies has allowed the accumulation of omics data at an alarming rate, thus introducing the era of 'big data'. Extracting inherent valuable knowledge from various omics data remains a daunting problem in bioinformatics. Better solutions often need some kind of more innovative methods for efficient handlings and effective results. Recent advancements in integrated analysis and computational modeling of multi-omics data helped address such needs in an increasingly harmonious manner. The development and application of machine learning have largely advanced our insights into biology and biomedicine and greatly promoted the development of therapeutic strategies, especially for precision medicine. Here, we propose a comprehensive survey and discussion on what happened, is happening and will happen when machine learning meets omics. Specifically, we describe how artificial intelligence can be applied to omics studies and review recent advancements at the interface between machine learning and the ever-widest range of omics including genomics, transcriptomics, proteomics, metabolomics, radiomics, as well as those at the single-cell resolution. We also discuss and provide a synthesis of ideas, new insights, current challenges and perspectives of machine learning in omics.

Leveraging the attention mechanism to improve the identification of DNA N6-methyladenine sites

DNA N6-methyladenine is an important type of DNA modification that plays important roles in multiple biological processes. Despite the recent progress in developing DNA 6mA site prediction methods, several challenges remain to be addressed. For example, although the hand-crafted features are interpretable, they contain redundant information that may bias the model training and have a negative impact on the trained model. Furthermore, although deep learning (DL)-based models can perform feature extraction and classification automatically, they lack the interpretability of the crucial features learned by those models. As such, considerable research efforts have been focused on achieving the trade-off between the interpretability and straightforwardness of DL neural networks. In this study, we develop two new DL-based models for improving the prediction of N6-methyladenine sites, termed LA6mA and AL6mA, which use bidirectional long short-term memory to respectively capture the long-range information and self-attention mechanism to extract the key position information from DNA sequences. The performance of the two proposed methods is benchmarked and evaluated on the two model organisms Arabidopsis thaliana and Drosophila melanogaster. On the two benchmark datasets, LA6mA achieves an area under the receiver operating characteristic curve (AUROC) value of 0.962 and 0.966, whereas AL6mA achieves an AUROC value of 0.945 and 0.941, respectively. Moreover, an in-depth analysis of the attention matrix is conducted to interpret the important information, which is hidden in the sequence and relevant for 6mA site prediction. The two novel pipelines developed for DNA 6mA site prediction in this work will facilitate a better understanding of the underlying principle of DL-based DNA methylation site prediction and its future applications.

Use of deep learning genomics to discriminate Alzheimer's disease and healthy controls

Alzheimer's disease (AD) is the most prevalent neurodegenerative disorder and the most common form of dementia in the elderly. Because gene is an important clinical risk factor resulting in AD, genomic studies, such as genome-wide association studies (GWAS), have widely been applied into AD studies. However, main shortcomings of GWAS method were that hereditary deletions were evident in the GWAS studies, which resulted in low classification or prediction abilities by using GWAS analysis. Therefore, this paper proposed a novel deep learning genomics approach and applied it to discriminate AD patients and healthy control (HC) subjects. In this study, we selected genotype data of 988 subjects enrolled in the ADNI, including 622 AD patients and 366 HC subjects. The proposed deep learning genomics (DLG) approach was composed of three steps: quality control, SNP genotype coding, and classification. The Resnet framework was used as the DLG model in this study. In the comparative GWAS analysis, APOE ε4 status and the normalized theta-value of the significant SNP loci were seen as predictors to classify genetically using Support Vector Machine (SVM). All data were divided into one training & validation group and one test group. 5-fold cross-validation was used in 500 times. Finally, we compared the classification results between DLG model and traditional GWAS analysis. As a result, the accuracy, sensitivity, and specificity of classification for traditional GWAS analysis was 71.38%±0.63%, 63.13%±2.87% and 85.59%±6.66% in the test group; while the accuracy, sensitivity, and specificity of classification for DLG model was 92.65%±4.80%, 85.00%±16.25% and 97.10%±4.38% in the test group. Hence, the DLG model can achieve higher accuracy and sensitivity when applied to AD. More importantly, we discovered several novel genetic biomarkers of AD, including rs6311 and rs6313 in HTR2A, and rs690705 in RFC3. The roles of these novel loci in AD should be explored future.

An inferred functional impact map of genetic variants in rice

Interpreting the functional impacts of genetic variants (GVs) is an important challenge for functional genomic studies in crops and next-generation breeding. Previous studies in rice (Oryza sativa) have focused mainly on the identification of GVs, whereas systematic functional annotation of GVs has not yet been performed. Here, we present a functional impact map of GVs in rice. We curated haplotype information for 17 397 026 GVs from sequencing data of 4726 rice accessions. We quantitatively evaluated the effects of missense mutations in coding regions in each haplotype based on the conservation of amino acid residues and obtained the effects of 918 848 non-redundant missense GVs. Furthermore, we generated high-quality chromatin accessibility (CA) data from six representative rice tissues and used these data to train deep convolutional neural network models to predict the impacts of 5 067 405 GVs for CA in regulatory regions. We characterized the functional properties and tissue specificity of the GV effects and found that large-effect GVs in coding and regulatory regions may be subject to selection in different directions. Finally, we demonstrated how the functional impact map could be used to prioritize causal variants in mapping populations. This impact map will be a useful resource for accelerating gene cloning and functional studies in rice, and can be freely queried in RiceVarMap V2.0 (http://ricevarmap.ncpgr.cn).

Discovering differential genome sequence activity with interpretable and efficient deep learning

Discovering sequence features that differentially direct cells to alternate fates is key to understanding both cellular development and the consequences of disease related mutations. We introduce Expected Pattern Effect and Differential Expected Pattern Effect, two black-box methods that can interpret genome regulatory sequences for cell type-specific or condition specific patterns. We show that these methods identify relevant transcription factor motifs and spacings that are predictive of cell state-specific chromatin accessibility. Finally, we integrate these methods into framework that is readily accessible to non-experts and available for download as a binary or installed via PyPI or bioconda at https://cgs.csail.mit.edu/deepaccess-package/.

Use of Deep-Learning Genomics to Discriminate Healthy Individuals from Those with Alzheimer's Disease or Mild Cognitive Impairment

OBJECTIVES

Alzheimer's disease (AD) is the most prevalent neurodegenerative disorder and the most common form of dementia in the elderly. Certain genes have been identified as important clinical risk factors for AD, and technological advances in genomic research, such as genome-wide association studies (GWAS), allow for analysis of polymorphisms and have been widely applied to studies of AD. However, shortcomings of GWAS include sensitivity to sample size and hereditary deletions, which result in low classification and predictive accuracy. Therefore, this paper proposes a novel deep-learning genomics approach and applies it to multitasking classification of AD progression, with the goal of identifying novel genetic biomarkers overlooked by traditional GWAS analysis.

METHODS

In this study, we selected genotype data from 1461 subjects enrolled in the Alzheimer's Disease Neuroimaging Initiative, including 622 AD, 473 mild cognitive impairment (MCI), and 366 healthy control (HC) subjects. The proposed deep-learning genomics (DLG) approach consists of three steps: quality control, coding of single-nucleotide polymorphisms, and classification. The ResNet framework was used for the DLG model, and the results were compared with classifications by simple convolutional neural network structure. All data were randomly assigned to one training/validation group and one test group at a ratio of 9 : 1. And fivefold cross-validation was used.

RESULTS

We compared classification results from the DLG model to those from traditional GWAS analysis among the three groups. For the AD and HC groups, the accuracy, sensitivity, and specificity of classification were, respectively, 98.78 ± 1.50%, 98.39% ± 2.50%, and 99.44% ± 1.11% using the DLG model, while 71.38% ± 0.63%, 63.13% ± 2.87%, and 85.59% ± 6.66% using traditional GWAS. Similar results were obtained from the other two intergroup classifications.

CONCLUSION

The DLG model can achieve higher accuracy and sensitivity when applied to progression of AD. More importantly, we discovered several novel genetic biomarkers of AD progression, including rs6311 and rs6313 in HTR2A, rs1354269 in NAV2, and rs690705 in RFC3. The roles of these novel loci in AD should be explored in future research.

Deciphering the genetic code of DNA methylation

DNA methylation plays crucial roles in many biological processes and abnormal DNA methylation patterns are often observed in diseases. Recent studies have shed light on cis-acting DNA elements that regulate locus-specific DNA methylation, which involves transcription factors, histone modification and DNA secondary structures. In addition, several recent studies have surveyed DNA motifs that regulate DNA methylation and suggest potential applications in diagnosis and prognosis. Here, we discuss the current biological foundation for the cis-acting genetic code that regulates DNA methylation. We review the computational models that predict DNA methylation with genetic features and discuss the biological insights revealed from these models. We also provide an in-depth discussion on how to leverage such knowledge in clinical applications, particularly in the context of liquid biopsy for early cancer diagnosis and treatment.

CRISPR technologies for precise epigenome editing

The epigenome involves a complex set of cellular processes governing genomic activity. Dissecting this complexity necessitates the development of tools capable of specifically manipulating these processes. The repurposing of prokaryotic CRISPR systems has allowed for the development of diverse technologies for epigenome engineering. Here, we review the state of currently achievable epigenetic manipulations along with corresponding applications. With future optimization, CRISPR-based epigenomic editing stands as a set of powerful tools for understanding and controlling biological function.

Overcoming Interpretability in Deep Learning Cancer Classification

Since its inception, deep learning has revolutionized the field of machine learning and data-driven science. One such data-driven science to be transformed by deep learning is genomics. In the past decade, numerous genomics studies have adopted deep learning and its applications range from predicting regulatory elements to cancer classification. Despite its dominating efficacy in these applications, deep learning is not without drawbacks. A prominent shortcoming of deep learning is the lack of interpretability. Hence, the main objective of this study is to address this obstacle in the deep learning cancer classification. Here we adopt a feature importance scoring methodology (Gradient-based class activation mapping or Grad-CAM) on a quasi-recurrent neural network model that classify cancer based on FASTA sequencing data. In this study, we managed to formulate a nucleotide-to-genomic-region Grad-CAM scoring methodology, as well as, validate the use this methodology for the chosen model. Consequently, this allows for the utilization of the Grad-CAM scoring methodology for feature importance in deep learning cancer classification. The results from our study identify potential novel candidate genes, genomic elements, and mechanisms for future cancer research.

Genomic Diversity in Sporadic Breast Cancer in a Latin American Population

Among Latin American women, breast cancer incidences vary across populations. Uruguay and Argentina have the highest rates in South America, which are mainly attributed to strong, genetic European contributions. Most genetic variants associated with breast cancer were described in European populations. However, the vast majority of genetic contributors to breast cancer risk remain unknown. Here, we report the results of a candidate gene association study of sporadic breast cancer in 176 cases and 183 controls in the Uruguayan population. We analyzed 141 variants from 98 loci that have been associated with overall breast cancer risk in European populations. We found weak evidence for the association of risk variants rs294174 (ESR1), rs16886165 (MAP3K1), rs2214681 (CNTNAP2), rs4237855 (VDR), rs9594579 (RANKL), rs8183919 (PTGIS), rs2981582 (FGFR2), and rs1799950 (BRCA1) with sporadic breast cancer. These results provide useful insight into the genetic susceptibility to sporadic breast cancer in the Uruguayan population and support the use of genetic risk scores for individualized screening and prevention.

RNA-Protein Binding Sites Prediction via Multi Scale Convolutional Gated Recurrent Unit Networks

RNA-Protein binding plays important roles in the field of gene expression. With the development of high throughput sequencing, several conventional methods and deep learning-based methods have been proposed to predict the binding preference of RNA-protein binding. These methods can hardly meet the need of consideration of the dependencies between subsequence and the various motif lengths of different translation factors (TFs). To overcome such limitations, we propose a predictive model that utilizes a combination of multi-scale convolutional layers and bidirectional gated recurrent unit (GRU) layer. Multi-scale convolution layer has the ability to capture the motif features of different lengths, and bidirectional GRU layer is able to capture the dependencies among subsequence. Experimental results show that the proposed method performs better than four state-of-the-art methods in this field. In addition, we investigate the effect of model structure on model performance by performing our proposed method with a different convolution layer and a different number of kernel size. We also demonstrate the effectiveness of bidirectional GRU in improving model performance through comparative experiments.

Predicting DNA Methylation States with Hybrid Information Based Deep-Learning Model

DNA methylation plays an important role in the regulation of some biological processes. Up to now, with the development of machine learning models, there are several sequence-based deep learning models designed to predict DNA methylation states, which gain better performance than traditional methods like random forest and SVM. However, convolutional network based deep learning models that use one-hot encoding DNA sequence as input may discover limited information and cause unsatisfactory prediction performance, so more data and model structures of diverse angles should be considered. In this work, we proposed a hybrid sequence-based deep learning model with both MeDIP-seq data and Histone information to predict DNA methylated CpG states (MHCpG). We combined both MeDIP-seq data and histone modification data with sequence information and implemented convolutional network to discover sequence patterns. In addition, we used statistical data gained from previous three input data and adopted a 3-layer feedforward neuron network to extract more high-level features. We compared our method with traditional predicting methods using random forest and other previous methods like CpGenie and DeepCpG, the result showed that MHCpG exceeded the other approaches and gained more satisfactory performance.

Prediction of RNA Methylation Status From Gene Expression Data Using Classification and Regression Methods

RNA N⁶-methyladenosine (m⁶A) has emerged as an important epigenetic modification for its role in regulating the stability, structure, processing, and translation of RNA. Instability of m⁶A homeostasis may result in flaws in stem cell regulation, decrease in fertility, and risk of cancer. To this day, experimental detection and quantification of RNA m⁶A modification are still time-consuming and labor-intensive. There is only a limited number of epitranscriptome samples in existing databases, and a matched RNA methylation profile is not often available for a biological problem of interests. As gene expression data are usually readily available for most biological problems, it could be appealing if we can estimate the RNA methylation status from gene expression data using in silico methods. In this study, we explored the possibility of computational prediction of RNA methylation status from gene expression data using classification and regression methods based on mouse RNA methylation data collected from 73 experimental conditions. Elastic Net-regularized Logistic Regression (ENLR), Support Vector Machine (SVM), and Random Forests (RF) were constructed for classification. Both SVM and RF achieved the best performance with the mean area under the curve (AUC) = 0.84 across samples; SVM had a narrower AUC spread. Gene Site Enrichment Analysis was conducted on those sites selected by ENLR as predictors to access the biological significance of the model. Three functional annotation terms were found statistically significant: phosphoprotein, SRC Homology 3 (SH3) domain, and endoplasmic reticulum. All 3 terms were found to be closely related to m⁶A pathway. For regression analysis, Elastic Net was implemented, which yielded a mean Pearson correlation coefficient = 0.68 and a mean Spearman correlation coefficient = 0.64. Our exploratory study suggested that gene expression data could be used to construct predictors for m⁶A methylation status with adequate accuracy. Our work showed for the first time that RNA methylation status may be predicted from the matched gene expression data. This finding may facilitate RNA modification research in various biological contexts when a matched RNA methylation profile is not available, especially in the very early stage of the study.

Amino acid encoding for deep learning applications

BACKGROUND

The number of applications of deep learning algorithms in bioinformatics is increasing as they usually achieve superior performance over classical approaches, especially, when bigger training datasets are available. In deep learning applications, discrete data, e.g. words or n-grams in language, or amino acids or nucleotides in bioinformatics, are generally represented as a continuous vector through an embedding matrix. Recently, learning this embedding matrix directly from the data as part of the continuous iteration of the model to optimize the target prediction - a process called 'end-to-end learning' - has led to state-of-the-art results in many fields. Although usage of embeddings is well described in the bioinformatics literature, the potential of end-to-end learning for single amino acids, as compared to more classical manually-curated encoding strategies, has not been systematically addressed. To this end, we compared classical encoding matrices, namely one-hot, VHSE8 and BLOSUM62, to end-to-end learning of amino acid embeddings for two different prediction tasks using three widely used architectures, namely recurrent neural networks (RNN), convolutional neural networks (CNN), and the hybrid CNN-RNN.

RESULTS

By using different deep learning architectures, we show that end-to-end learning is on par with classical encodings for embeddings of the same dimension even when limited training data is available, and might allow for a reduction in the embedding dimension without performance loss, which is critical when deploying the models to devices with limited computational capacities. We found that the embedding dimension is a major factor in controlling the model performance. Surprisingly, we observed that deep learning models are capable of learning from random vectors of appropriate dimension.

CONCLUSION

Our study shows that end-to-end learning is a flexible and powerful method for amino acid encoding. Further, due to the flexibility of deep learning systems, amino acid encoding schemes should be benchmarked against random vectors of the same dimension to disentangle the information content provided by the encoding scheme from the distinguishability effect provided by the scheme.

SRHiC: A Deep Learning Model to Enhance the Resolution of Hi-C Data

Hi-C data is important for studying chromatin three-dimensional structure. However, the resolution of most existing Hi-C data is generally coarse due to sequencing cost. Therefore, it will be helpful if we can predict high-resolution Hi-C data from low-coverage sequencing data. Here we developed a novel and simple computational method based on deep learning named super-resolution Hi-C (SRHiC) to enhance the resolution of Hi-C data. We verified SRHiC on Hi-C data in human cell line. We also evaluated the generalization power of SRHiC by enhancing Hi-C data resolution in other human and mouse cell types. Results showed that SRHiC outperforms the state-of-the-art methods in accuracy of prediction.

DeepMILO: a deep learning approach to predict the impact of non-coding sequence variants on 3D chromatin structure

Non-coding variants have been shown to be related to disease by alteration of 3D genome structures. We propose a deep learning method, DeepMILO, to predict the effects of variants on CTCF/cohesin-mediated insulator loops. Application of DeepMILO on variants from whole-genome sequences of 1834 patients of twelve cancer types revealed 672 insulator loops disrupted in at least 10% of patients. Our results show mutations at loop anchors are associated with upregulation of the cancer driver genes BCL2 and MYC in malignant lymphoma thus pointing to a possible new mechanism for their dysregulation via alteration of insulator loops.

MethylNet: an automated and modular deep learning approach for DNA methylation analysis

BACKGROUND

DNA methylation (DNAm) is an epigenetic regulator of gene expression programs that can be altered by environmental exposures, aging, and in pathogenesis. Traditional analyses that associate DNAm alterations with phenotypes suffer from multiple hypothesis testing and multi-collinearity due to the high-dimensional, continuous, interacting and non-linear nature of the data. Deep learning analyses have shown much promise to study disease heterogeneity. DNAm deep learning approaches have not yet been formalized into user-friendly frameworks for execution, training, and interpreting models. Here, we describe MethylNet, a DNAm deep learning method that can construct embeddings, make predictions, generate new data, and uncover unknown heterogeneity with minimal user supervision.

RESULTS

The results of our experiments indicate that MethylNet can study cellular differences, grasp higher order information of cancer sub-types, estimate age and capture factors associated with smoking in concordance with known differences.

CONCLUSION

The ability of MethylNet to capture nonlinear interactions presents an opportunity for further study of unknown disease, cellular heterogeneity and aging processes.

Identifying viruses from metagenomic data using deep learning

BACKGROUND

The recent development of metagenomic sequencing makes it possible to massively sequence microbial genomes including viral genomes without the need for laboratory culture. Existing reference-based and gene homology-based methods are not efficient in identifying unknown viruses or short viral sequences from metagenomic data.

METHODS

Here we developed a reference-free and alignment-free machine learning method, DeepVirFinder, for identifying viral sequences in metagenomic data using deep learning.

RESULTS

Trained based on sequences from viral RefSeq discovered before May 2015, and evaluated on those discovered after that date, DeepVirFinder outperformed the state-of-the-art method VirFinder at all contig lengths, achieving AUROC 0.93, 0.95, 0.97, and 0.98 for 300, 500, 1000, and 3000 bp sequences respectively. Enlarging the training data with additional millions of purified viral sequences from metavirome samples further improved the accuracy for identifying virus groups that are under-represented. Applying DeepVirFinder to real human gut metagenomic samples, we identified 51,138 viral sequences belonging to 175 bins in patients with colorectal carcinoma (CRC). Ten bins were found associated with the cancer status, suggesting viruses may play important roles in CRC.

CONCLUSIONS

Powered by deep learning and high throughput sequencing metagenomic data, DeepVirFinder significantly improved the accuracy of viral identification and will assist the study of viruses in the era of metagenomics.

Being Born Large for Gestational Age is Associated with Increased Global Placental DNA Methylation

Being born small (SGA) or large for gestational age (LGA) is associated with adverse birth outcomes and metabolic diseases in later life of the offspring. It is known that aberrations in growth during gestation are related to altered placental function. Placental function is regulated by epigenetic mechanisms such as DNA methylation. Several studies in recent years have demonstrated associations between altered patterns of DNA methylation and adverse birth outcomes. However, larger studies that reliably investigated global DNA methylation are lacking. The aim of this study was to characterize global placental DNA methylation in relationship to size for gestational age. Global DNA methylation was assessed in 1023 placental samples by LC-MS/MS. LGA offspring displayed significantly higher global placental DNA methylation compared to appropriate for gestational age (AGA; p < 0.001). ANCOVA analyses adjusted for known factors impacting on DNA methylation demonstrated an independent association between placental global DNA methylation and LGA births (p < 0.001). Tertile stratification according to global placental DNA methylation levels revealed a significantly higher frequency of LGA births in the third tertile. Furthermore, a multiple logistic regression analysis corrected for known factors influencing birth weight highlighted an independent positive association between global placental DNA methylation and the frequency of LGA births (p = 0.001).

A comprehensive review of computational prediction of genome-wide features

There are significant correlations among different types of genetic, genomic and epigenomic features within the genome. These correlations make the in silico feature prediction possible through statistical or machine learning models. With the accumulation of a vast amount of high-throughput data, feature prediction has gained significant interest lately, and a plethora of papers have been published in the past few years. Here we provide a comprehensive review on these published works, categorized by the prediction targets, including protein binding site, enhancer, DNA methylation, chromatin structure and gene expression. We also provide discussions on some important points and possible future directions.

Deep learning in omics: a survey and guideline

Omics, such as genomics, transcriptome and proteomics, has been affected by the era of big data. A huge amount of high dimensional and complex structured data has made it no longer applicable for conventional machine learning algorithms. Fortunately, deep learning technology can contribute toward resolving these challenges. There is evidence that deep learning can handle omics data well and resolve omics problems. This survey aims to provide an entry-level guideline for researchers, to understand and use deep learning in order to solve omics problems. We first introduce several deep learning models and then discuss several research areas which have combined omics and deep learning in recent years. In addition, we summarize the general steps involved in using deep learning which have not yet been systematically discussed in the existent literature on this topic. Finally, we compare the features and performance of current mainstream open source deep learning frameworks and present the opportunities and challenges involved in deep learning. This survey will be a good starting point and guideline for omics researchers to understand deep learning.

Deciphering epigenomic code for cell differentiation using deep learning

Background

Although DNA sequence plays a crucial role in establishing the unique epigenome of a cell type, little is known about the sequence determinants that lead to the unique epigenomes of different cell types produced during cell differentiation. To fill this gap, we employed two types of deep convolutional neural networks (CNNs) constructed for each of differentially related cell types and for each of histone marks measured in the cells, to learn the sequence determinants of various histone modification patterns in each cell type.

Results

We applied our models to four differentially related human CD₄⁺ T cell types and six histone marks measured in each cell type. The cell models can accurately predict the histone marks in each cell type, while the mark models can also accurately predict the cell types based on a single mark. Sequence motifs learned by both the cell or mark models are highly similar to known binding motifs of transcription factors known to play important roles in CD₄⁺ T cell differentiation. Both the unique histone mark patterns in each cell type and the different patterns of the same histone mark in different cell types are determined by a set of motifs with unique combinations. Interestingly, the level of sharing motifs learned in the different cell models reflects the lineage relationships of the cells, while the level of sharing motifs learned in the different histone mark models reflects their functional relationships. These models can also enable the prediction of the importance of learned motifs and their interactions in determining specific histone mark patterns in the cell types.

Conclusion

Sequence determinants of various histone modification patterns in different cell types can be revealed by comparative analysis of motifs learned in the CNN models for multiple cell types and histone marks. The learned motifs are interpretable and may provide insights into the underlying molecular mechanisms of establishing the unique epigenomes in different cell types. Thus, our results support the hypothesis that DNA sequences ultimately determine the unique epigenomes of different cell types through their interactions with transcriptional factors, epigenome remodeling system and extracellular cues during cell differentiation.

Electronic supplementary material

The online version of this article (10.1186/s12864-019-6072-8) contains supplementary material, which is available to authorized users.

Identification of DNA motifs that regulate DNA methylation

DNA methylation is an important epigenetic mark but how its locus-specificity is decided in relation to DNA sequence is not fully understood. Here, we have analyzed 34 diverse whole-genome bisulfite sequencing datasets in human and identified 313 motifs, including 92 and 221 associated with methylation (methylation motifs, MMs) and unmethylation (unmethylation motifs, UMs), respectively. The functionality of these motifs is supported by multiple lines of evidence. First, the methylation levels at the MM and UM motifs are respectively higher and lower than the genomic background. Second, these motifs are enriched at the binding sites of methylation modifying enzymes including DNMT3A and TET1, indicating their possible roles of recruiting these enzymes. Third, these motifs significantly overlap with "somatic QTLs" (quantitative trait loci) of methylation and expression. Fourth, disruption of these motifs by mutation is associated with significantly altered methylation level of the CpGs in the neighbor regions. Furthermore, these motifs together with somatic mutations are predictive of cancer subtypes and patient survival. We revealed some of these motifs were also associated with histone modifications, suggesting a possible interplay between the two types of epigenetic modifications. We also found some motifs form feed forward loops to contribute to DNA methylation dynamics.

Deep learning: new computational modelling techniques for genomics

As a data-driven science, genomics largely utilizes machine learning to capture dependencies in data and derive novel biological hypotheses. However, the ability to extract new insights from the exponentially increasing volume of genomics data requires more expressive machine learning models. By effectively leveraging large data sets, deep learning has transformed fields such as computer vision and natural language processing. Now, it is becoming the method of choice for many genomics modelling tasks, including predicting the impact of genetic variation on gene regulatory mechanisms such as DNA accessibility and splicing.

Visualizing complex feature interactions and feature sharing in genomic deep neural networks

BACKGROUND

Visualization tools for deep learning models typically focus on discovering key input features without considering how such low level features are combined in intermediate layers to make decisions. Moreover, many of these methods examine a network's response to specific input examples that may be insufficient to reveal the complexity of model decision making.

RESULTS

We present DeepResolve, an analysis framework for deep convolutional models of genome function that visualizes how input features contribute individually and combinatorially to network decisions. Unlike other methods, DeepResolve does not depend upon the analysis of a predefined set of inputs. Rather, it uses gradient ascent to stochastically explore intermediate feature maps to 1) discover important features, 2) visualize their contribution and interaction patterns, and 3) analyze feature sharing across tasks that suggests shared biological mechanism. We demonstrate the visualization of decision making using our proposed method on deep neural networks trained on both experimental and synthetic data. DeepResolve is competitive with existing visualization tools in discovering key sequence features, and identifies certain negative features and non-additive feature interactions that are not easily observed with existing tools. It also recovers similarities between poorly correlated classes which are not observed by traditional methods. DeepResolve reveals that DeepSEA's learned decision structure is shared across genome annotations including histone marks, DNase hypersensitivity, and transcription factor binding. We identify groups of TFs that suggest known shared biological mechanism, and recover correlation between DNA hypersensitivities and TF/Chromatin marks.

CONCLUSIONS

DeepResolve is capable of visualizing complex feature contribution patterns and feature interactions that contribute to decision making in genomic deep convolutional networks. It also recovers feature sharing and class similarities which suggest interesting biological mechanisms. DeepResolve is compatible with existing visualization tools and provides complementary insights.

LightCpG: a multi-view CpG sites detection on single-cell whole genome sequence data

BACKGROUND

DNA methylation plays an important role in multiple biological processes that are closely related to human health. The study of DNA methylation can provide an insight into the mechanism behind human health and can also have a positive effect on the assessment of human health status. However, the available sequencing technology is limited by incomplete CpG coverage. Therefore, it is crucial to discover an efficient and convenient method capable of distinguishing between the states of CpG sites. Previous studies focused on identifying methylation states of the CpG sites in single cell, which only evaluated sequence information or structural information.

RESULTS

In this paper, we propose a novel model, LightCpG, which combines the positional features with the sequence and structural features to provide information on the CpG sites at two stages. Next, we used the LightGBM model for training of the CpG site identification, and further utilized sample extraction and merged features to reduce the training time. Our results indicate that our method achieves outstanding performance in recognition of DNA methylation. The average AUC values of our method using the 25 human hepatocellular carcinoma cells (HCC) cell datasets and six human heptoplastoma-derived (HepG2) cell datasets were 0.9616 and 0.9213, respectively. Moreover, the average training times for our method on the HCC and HepG2 datasets were 8.3 and 5.06 s, respectively. Furthermore, the computational complexity of our model was much lower compared with other available methods that detect methylation states of the CpG sites.

CONCLUSIONS

In summary, LightCpG is an accurate model for identifying the DNA methylation status of CpG sites in single cells. Furthermore, three types of feature extraction methods and two strategies used in LightCpG are helpful for other prediction problems.

Deep Learning in the Biomedical Applications: Recent and Future Status

Deep neural networks represent, nowadays, the most effective machine learning technology in biomedical domain. In this domain, the different areas of interest concern the Omics (study of the genome—genomics—and proteins—transcriptomics, proteomics, and metabolomics), bioimaging (study of biological cell and tissue), medical imaging (study of the human organs by creating visual representations), BBMI (study of the brain and body machine interface) and public and medical health management (PmHM). This paper reviews the major deep learning concepts pertinent to such biomedical applications. Concise overviews are provided for the Omics and the BBMI. We end our analysis with a critical discussion, interpretation and relevant open challenges.

Locus-specific DNA methylation prediction in cord blood and placenta

DNA methylation is known to be responsive to prenatal exposures, which may be a part of the mechanism linking early developmental exposures to future chronic diseases. Many studies use blood to measure DNA methylation, yet we know that DNA methylation is tissue specific. Placenta is central to fetal growth and development, but it is rarely feasible to collect this tissue in large epidemiological studies; on the other hand, cord blood samples are more accessible. In this study, based on paired samples of both placenta and cord blood tissues from 169 individuals, we investigated the methylation concordance between placenta and cord blood. We then employed a machine-learning-based model to predict locus-specific DNA methylation levels in placenta using DNA methylation levels in cord blood. We found that methylation correlation between placenta and cord blood is lower than other tissue pairs, consistent with existing observations that placenta methylation has a distinct pattern. Nonetheless, there are still a number of CpG sites showing robust association between the two tissues. We built prediction models for placenta methylation based on cord blood data and documented a subset of 1,012 CpG sites with high correlation between measured and predicted placenta methylation levels. The resulting list of CpG sites and prediction models could help to reveal the loci where internal or external influences may affect DNA methylation in both placenta and cord blood, and provide a reference data to predict the effects on placenta in future study even when the tissue is not available in an epidemiological study.

Allele-specific epigenome maps reveal sequence-dependent stochastic switching at regulatory loci

To assess the impact of genetic variation in regulatory loci on human health, we constructed a high-resolution map of allelic imbalances in DNA methylation, histone marks, and gene transcription in 71 epigenomes from 36 distinct cell and tissue types from 13 donors. Deep whole-genome bisulfite sequencing of 49 methylomes revealed sequence-dependent CpG methylation imbalances at thousands of heterozygous regulatory loci. Such loci are enriched for stochastic switching, which is defined as random transitions between fully methylated and unmethylated states of DNA. The methylation imbalances at thousands of loci are explainable by different relative frequencies of the methylated and unmethylated states for the two alleles. Further analyses provided a unifying model that links sequence-dependent allelic imbalances of the epigenome, stochastic switching at gene regulatory loci, and disease-associated genetic variation.

BoostMe accurately predicts DNA methylation values in whole-genome bisulfite sequencing of multiple human tissues

Background

Bisulfite sequencing is widely employed to study the role of DNA methylation in disease; however, the data suffer from biases due to coverage depth variability. Imputation of methylation values at low-coverage sites may mitigate these biases while also identifying important genomic features associated with predictive power.

Results

Here we describe BoostMe, a method for imputing low-quality DNA methylation estimates within whole-genome bisulfite sequencing (WGBS) data. BoostMe uses a gradient boosting algorithm, XGBoost, and leverages information from multiple samples for prediction. We find that BoostMe outperforms existing algorithms in speed and accuracy when applied to WGBS of human tissues. Furthermore, we show that imputation improves concordance between WGBS and the MethylationEPIC array at low WGBS depth, suggesting improved WGBS accuracy after imputation.

Conclusions

Our findings support the use of BoostMe as a preprocessing step for WGBS analysis.

Electronic supplementary material

The online version of this article (10.1186/s12864-018-4766-y) contains supplementary material, which is available to authorized users.