Deep learning for DNase I hypersensitive sites identification

Background

The DNase I hypersensitive sites (DHSs) are associated with the cis-regulatory DNA elements. An efficient method of identifying DHSs can enhance the understanding on the accessibility of chromatin. Despite a multitude of resources available on line including experimental datasets and computational tools, the complex language of DHSs remains incompletely understood.

Methods

Here, we address this challenge using an approach based on a state-of-the-art machine learning method. We present a novel convolutional neural network (CNN) which combined Inception like networks with a gating mechanism for the response of multiple patterns and longterm association in DNA sequences to predict multi-scale DHSs in Arabidopsis, rice and Homo sapiens.

Results

Our method obtains 0.961 area under curve (AUC) on Arabidopsis, 0.969 AUC on rice and 0.918 AUC on Homo sapiens.

Conclusions

Our method provides an efficient and accurate way to identify multi-scale DHSs sequences by deep learning.

     

IGENT: efficient entropy based algorithm for genome-wide gene-gene interaction analysis

Background

With the development of high-throughput genotyping and sequencing technology, there are growing evidences of association with genetic variants and complex traits. In spite of thousands of genetic variants discovered, such genetic markers have been shown to explain only a very small proportion of the underlying genetic variance of complex traits. Gene-gene interaction (GGI) analysis is expected to unveil a large portion of unexplained heritability of complex traits.

Methods

In this work, we propose IGENT, Information theory-based GEnome-wide gene-gene iNTeraction method. IGENT is an efficient algorithm for identifying genome-wide gene-gene interactions (GGI) and gene-environment interaction (GEI). For detecting significant GGIs in genome-wide scale, it is important to reduce computational burden significantly. Our method uses information gain (IG) and evaluates its significance without resampling.

Results

Through our simulation studies, the power of the IGENT is shown to be better than or equivalent to that of that of BOOST. The proposed method successfully detected GGI for bipolar disorder in the Wellcome Trust Case Control Consortium (WTCCC) and age-related macular degeneration (AMD).

Conclusions

The proposed method is implemented by C++ and available on Windows, Linux and MacOSX.

     

GxGrare: gene-gene interaction analysis method for rare variants from high-throughput sequencing data

Background

With the rapid advancement of array-based genotyping techniques, genome-wide association studies (GWAS) have successfully identified common genetic variants associated with common complex diseases. However, it has been shown that only a small proportion of the genetic etiology of complex diseases could be explained by the genetic factors identified from GWAS. This missing heritability could possibly be explained by gene-gene interaction (epistasis) and rare variants. There has been an exponential growth of gene-gene interaction analysis for common variants in terms of methodological developments and practical applications. Also, the recent advancement of high-throughput sequencing technologies makes it possible to conduct rare variant analysis. However, little progress has been made in gene-gene interaction analysis for rare variants.

Results

Here, we propose GxGrare which is a new gene-gene interaction method for the rare variants in the framework of the multifactor dimensionality reduction (MDR) analysis. The proposed method consists of three steps; 1) collapsing the rare variants, 2) MDR analysis for the collapsed rare variants, and 3) detect top candidate interaction pairs. GxGrare can be used for the detection of not only gene-gene interactions, but also interactions within a single gene. The proposed method is illustrated with 1080 whole exome sequencing data of the Korean population in order to identify causal gene-gene interaction for rare variants for type 2 diabetes.

Conclusion

The proposed GxGrare performs well for gene-gene interaction detection with collapsing of rare variants. GxGrare is available at http://bibs.snu.ac.kr/software/gxgrare which contains simulation data and documentation. Supported operating systems include Linux and OS X.

     

Detection of potential new biomarkers of atherosclerosis by probe electrospray ionization mass spectrometry

Introduction

Atherosclerotic diseases are the leading cause of death worldwide. Biomarkers of atherosclerosis are required to monitor and prevent disease progression. While mass spectrometry is a promising technique to search for such biomarkers, its clinical application is hampered by the laborious processes for sample preparation and analysis.

Methods

We developed a rapid method to detect plasma metabolites by probe electrospray ionization mass spectrometry (PESI-MS), which employs an ambient ionization technique enabling atmospheric pressure rapid mass spectrometry. To create an automatic diagnosis system of atherosclerotic disorders, we applied machine learning techniques to the obtained spectra.

Results

Using our system, we successfully discriminated between rabbits with and without dyslipidemia. The causes of dyslipidemia (genetic lipoprotein receptor deficiency or dietary cholesterol overload) were also distinguishable by this method. Furthermore, after induction of atherosclerosis in rabbits with a cholesterol-rich diet, we were able to detect dynamic changes in plasma metabolites. The major metabolites detected by PESI-MS included cholesterol sulfate and a phospholipid (PE18:0/20:4), which are promising new biomarkers of atherosclerosis.

Conclusion

We developed a remarkably fast and easy method to detect potential new biomarkers of atherosclerosis in plasma using PESI-MS.

     

Predicting adverse drug reactions through interpretable deep learning framework

Background

Adverse drug reactions (ADRs) are unintended and harmful reactions caused by normal uses of drugs. Predicting and preventing ADRs in the early stage of the drug development pipeline can help to enhance drug safety and reduce financial costs.

Methods

In this paper, we developed machine learning models including a deep learning framework which can simultaneously predict ADRs and identify the molecular substructures associated with those ADRs without defining the substructures a-priori.

Results

We evaluated the performance of our model with ten different state-of-the-art fingerprint models and found that neural fingerprints from the deep learning model outperformed all other methods in predicting ADRs. Via feature analysis on drug structures, we identified important molecular substructures that are associated with specific ADRs and assessed their associations via statistical analysis.

Conclusions

The deep learning model with feature analysis, substructure identification, and statistical assessment provides a promising solution for identifying risky components within molecular structures and can potentially help to improve drug safety evaluation.

     

Integrative pathway-based survival prediction utilizing the interaction between gene expression and DNA methylation in breast cancer

Background

Integrative analysis on multi-omics data has gained much attention recently. To investigate the interactive effect of gene expression and DNA methylation on cancer, we propose a directed random walk-based approach on an integrated gene-gene graph that is guided by pathway information.

Methods

Our approach first extracts a single pathway profile matrix out of the gene expression and DNA methylation data by performing the random walk over the integrated graph. We then apply a denoising autoencoder to the pathway profile to further identify important pathway features and genes. The extracted features are validated in the survival prediction task for breast cancer patients.

Results

The results show that the proposed method substantially improves the survival prediction performance compared to that of other pathway-based prediction methods, revealing that the combined effect of gene expression and methylation data is well reflected in the integrated gene-gene graph combined with pathway information. Furthermore, we show that our joint analysis on the methylation features and gene expression profile identifies cancer-specific pathways with genes related to breast cancer.

Conclusions

In this study, we proposed a DRW-based method on an integrated gene-gene graph with expression and methylation profiles in order to utilize the interactions between them. The results showed that the constructed integrated gene-gene graph can successfully reflect the combined effect of methylation features on gene expression profiles. We also found that the selected features by DA can effectively extract topologically important pathways and genes specifically related to breast cancer.

     

Data mining and machine learning approaches for the integration of genome-wide association and methylation data: methodology and main conclusions from GAW20

Background

GAW20 working group 5 brought together researchers who contributed 7 papers with the aim of evaluating methods to detect genetic by epigenetic interactions. GAW20 distributed real data from the Genetics of Lipid Lowering Drugs and Diet Network (GOLDN) study, including single-nucleotide polymorphism (SNP) markers, methylation (cytosine-phosphate-guanine [CpG]) markers, and phenotype information on up to 995 individuals. In addition, a simulated data set based on the real data was provided.

Results

The 7 contributed papers analyzed these data sets with a number of different statistical methods, including generalized linear mixed models, mediation analysis, machine learning, W-test, and sparsity-inducing regularized regression. These methods generally appeared to perform well. Several papers confirmed a number of causative SNPs in either the large number of simulation sets or the real data on chromosome 11. Findings were also reported for different SNPs, CpG sites, and SNP–CpG site interaction pairs.

Conclusions

In the simulation (200 replications), power appeared generally good for large interaction effects, but smaller effects will require larger studies or consortium collaboration for realizing a sufficient power.

     

Semi-supervised drug-protein interaction prediction from heterogeneous biological spaces

Background

Predicting drug-protein interactions from heterogeneous biological data sources is a key step for in silico drug discovery. The difficulty of this prediction task lies in the rarity of known drug-protein interactions and myriad unknown interactions to be predicted. To meet this challenge, a manifold regularization semi-supervised learning method is presented to tackle this issue by using labeled and unlabeled information which often generates better results than using the labeled data alone. Furthermore, our semi-supervised learning method integrates known drug-protein interaction network information as well as chemical structure and genomic sequence data.

Results

Using the proposed method, we predicted certain drug-protein interactions on the enzyme, ion channel, GPCRs, and nuclear receptor data sets. Some of them are confirmed by the latest publicly available drug targets databases such as KEGG.

Conclusions

We report encouraging results of using our method for drug-protein interaction network reconstruction which may shed light on the molecular interaction inference and new uses of marketed drugs.

     

Biomedical event trigger detection by dependency-based word embedding

Background

In biomedical research, events revealing complex relations between entities play an important role. Biomedical event trigger identification has become a research hotspot since its important role in biomedical event extraction. Traditional machine learning methods, such as support vector machines (SVM) and maxent classifiers, which aim to manually design powerful features fed to the classifiers, depend on the understanding of the specific task and cannot generalize to the new domain or new examples.

Methods

In this paper, we propose an approach which utilizes neural network model based on dependency-based word embedding to automatically learn significant features from raw input for trigger classification. First, we employ Word2vecf, the modified version of Word2vec, to learn word embedding with rich semantic and functional information based on dependency relation tree. Then neural network architecture is used to learn more significant feature representation based on raw dependency-based word embedding. Meanwhile, we dynamically adjust the embedding while training for adapting to the trigger classification task. Finally, softmax classifier labels the examples by specific trigger class using the features learned by the model.

Results

The experimental results show that our approach achieves a micro-averaging F1 score of 78.27 and a macro-averaging F1 score of 76.94 % in significant trigger classes, and performs better than baseline methods. In addition, we can achieve the semantic distributed representation of every trigger word.

     

Feed-forward neural network model for hunger and satiety related VAS score prediction

Background

An artificial neural network approach was chosen to model the outcome of the complex signaling pathways in the gastro-intestinal tract and other peripheral organs that eventually produce the satiety feeling in the brain upon feeding.

Methods

A multilayer feed-forward neural network was trained with sets of experimental data relating concentration-time courses of plasma satiety hormones to Visual Analog Scales (VAS) scores. The network successfully predicted VAS responses from sets of satiety hormone data obtained in experiments using different food compositions.

Results

The correlation coefficients for the predicted VAS responses for test sets having i) a full set of three satiety hormones, ii) a set of only two satiety hormones, and iii) a set of only one satiety hormone were 0.96, 0.96, and 0.89, respectively. The predicted VAS responses discriminated the satiety effects of high satiating food types from less satiating food types both in orally fed and ileal infused forms.

Conclusions

From this application of artificial neural networks, one may conclude that neural network models are very suitable to describe situations where behavior is complex and incompletely understood. However, training data sets that fit the experimental conditions need to be available.

     

Ancestors’ dietary patterns and environments could drive positive selection in genes involved in micronutrient metabolism—the case of cofactor transporters

Background

During evolution, humans colonized different ecological niches and adopted a variety of subsistence strategies that gave rise to diverse selective pressures acting across the genome. Environmentally induced selection of vitamin, mineral, or other cofactor transporters could influence micronutrient-requiring molecular reactions and contribute to inter-individual variability in response to foods and nutritional interventions.

Methods

A comprehensive list of genes coding for transporters of cofactors or their precursors was built using data mining procedures from the HGDP dataset and then explored to detect evidence of positive genetic selection. This dataset was chosen since it comprises several genetically diverse worldwide populations whom ancestries have evolved in different environments and thus lived following various nutritional habits and lifestyles.

Results

We identified 312 cofactor transporter (CT) genes involved in between-cell or sub-cellular compartment distribution of 28 cofactors derived from dietary intake. Twenty-four SNPs distributed across 14 CT genes separated populations into continental and intra-continental groups such as African hunter-gatherers and farmers, and between Native American sub-populations. Notably, four SNPs were located in SLC24A3 with one being a known eQTL of the NCKX3 protein.

Conclusions

These findings could support the importance of considering individual’s genetic makeup along with their metabolic profile when tailoring personalized dietary interventions for optimizing health.

     

Protein complex detection with semi-supervised learning in protein interaction networks

Background

Protein-protein interactions (PPIs) play fundamental roles in nearly all biological processes. The systematic analysis of PPI networks can enable a great understanding of cellular organization, processes and function. In this paper, we investigate the problem of protein complex detection from noisy protein interaction data, i.e., finding the subsets of proteins that are closely coupled via protein interactions. However, protein complexes are likely to overlap and the interaction data are very noisy. It is a great challenge to effectively analyze the massive data for biologically meaningful protein complex detection.

Results

Many people try to solve the problem by using the traditional unsupervised graph clustering methods. Here, we stand from a different point of view, redefining the properties and features for protein complexes and designing a “semi-supervised” method to analyze the problem. In this paper, we utilize the neural network with the “semi-supervised” mechanism to detect the protein complexes. By retraining the neural network model recursively, we could find the optimized parameters for the model, in such a way we can successfully detect the protein complexes. The comparison results show that our algorithm could identify protein complexes that are missed by other methods. We also have shown that our method achieve better precision and recall rates for the identified protein complexes than other existing methods. In addition, the framework we proposed is easy to be extended in the future.

Conclusions

Using a weighted network to represent the protein interaction network is more appropriate than using a traditional unweighted network. In addition, integrating biological features and topological features to represent protein complexes is more meaningful than using dense subgraphs. Last, the “semi-supervised” learning model is a promising model to detect protein complexes with more biological and topological features available.

     

Predicting synthetic lethal genetic interactions in <Emphasis Type="Italic">Saccharomyces cerevisiae</Emphasis> using short polypeptide clusters

Background

Protein synthetic lethal genetic interactions are useful to define functional relationships between proteins and pathways. However, the molecular mechanism of synthetic lethal genetic interactions remains unclear.

Results

In this study we used the clusters of short polypeptide sequences, which are typically shorter than the classically defined protein domains, to characterize the functionalities of proteins. We developed a framework to identify significant short polypeptide clusters from yeast protein sequences, and then used these short polypeptide clusters as features to predict yeast synthetic lethal genetic interactions. The short polypeptide clusters based approach provides much higher coverage for predicting yeast synthetic lethal genetic interactions. Evaluation using experimental data sets showed that the short polypeptide clusters based approach is superior to the previous protein domain based one.

Conclusion

We were able to achieve higher performance in yeast synthetic lethal genetic interactions prediction using short polypeptide clusters as features. Our study suggests that the short polypeptide cluster may help better understand the functionalities of proteins.

     

Chaperone–substrate interactions monitored via a robust TEM-1 β-lactamase fragment complementation assay

Objective

To investigate the application of the TEM-1 β-lactamase protein fragment complementation assay (PCA) in detecting weak and unstable protein–protein interactions as typically observed during chaperone-assisted protein folding in the periplasm of Escherichia coli.

Results

The TEM-1 β-lactamase PCA system effectively captured the interactions of three pairs of chaperones and substrates. Moreover, the strength of the interactions can be quantitatively analyzed by comparing different levels of penicillin resistance, and the assay can be performed under 0.5% butanol, a stress condition thought to be physiologically relevant.

Conclusions

The β-lactamase PCA system faithfully reports chaperone-substrate interactions in the bacterial cell envelope, and therefore this system has the potential to map the complex protein homeostasis network under a fluctuating environment.

     

Predicting enhancers with deep convolutional neural networks

Background

With the rapid development of deep sequencing techniques in the recent years, enhancers have been systematically identified in such projects as FANTOM and ENCODE, forming genome-wide landscapes in a series of human cell lines. Nevertheless, experimental approaches are still costly and time consuming for large scale identification of enhancers across a variety of tissues under different disease status, making computational identification of enhancers indispensable.

Results

To facilitate the identification of enhancers, we propose a computational framework, named DeepEnhancer, to distinguish enhancers from background genomic sequences. Our method purely relies on DNA sequences to predict enhancers in an end-to-end manner by using a deep convolutional neural network (CNN). We train our deep learning model on permissive enhancers and then adopt a transfer learning strategy to fine-tune the model on enhancers specific to a cell line. Results demonstrate the effectiveness and efficiency of our method in the classification of enhancers against random sequences, exhibiting advantages of deep learning over traditional sequence-based classifiers. We then construct a variety of neural networks with different architectures and show the usefulness of such techniques as max-pooling and batch normalization in our method. To gain the interpretability of our approach, we further visualize convolutional kernels as sequence logos and successfully identify similar motifs in the JASPAR database.

Conclusions

DeepEnhancer enables the identification of novel enhancers using only DNA sequences via a highly accurate deep learning model. The proposed computational framework can also be applied to similar problems, thereby prompting the use of machine learning methods in life sciences.

     

Using multitask classification methods to investigate the kinase-specific phosphorylation sites

Background

Identification of phosphorylation sites by computational methods is becoming increasingly important because it reduces labor-intensive and costly experiments and can improve our understanding of the common properties and underlying mechanisms of protein phosphorylation.

Methods

A multitask learning framework for learning four kinase families simultaneously, instead of studying each kinase family of phosphorylation sites separately, is presented in the study. The framework includes two multitask classification methods: the Multi-Task Least Squares Support Vector Machines (MTLS-SVMs) and the Multi-Task Feature Selection (MT-Feat3).

Results

Using the multitask learning framework, we successfully identify 18 common features shared by four kinase families of phosphorylation sites. The reliability of selected features is demonstrated by the consistent performance in two multi-task learning methods.

Conclusions

The selected features can be used to build efficient multitask classifiers with good performance, suggesting they are important to protein phosphorylation across 4 kinase families.

     

Derivative-free neural network for optimizing the scoring functions associated with dynamic programming of pairwise-profile alignment

Background

A profile-comparison method with position-specific scoring matrix (PSSM) is among the most accurate alignment methods. Currently, cosine similarity and correlation coefficients are used as scoring functions of dynamic programming to calculate similarity between PSSMs. However, it is unclear whether these functions are optimal for profile alignment methods. By definition, these functions cannot capture nonlinear relationships between profiles. Therefore, we attempted to discover a novel scoring function, which was more suitable for the profile-comparison method than existing functions, using neural networks.

Results

Although neural networks required derivative-of-cost functions, the problem being addressed in this study lacked them. Therefore, we implemented a novel derivative-free neural network by combining a conventional neural network with an evolutionary strategy optimization method used as a solver. Using this novel neural network system, we optimized the scoring function to align remote sequence pairs. Our results showed that the pairwise-profile aligner using the novel scoring function significantly improved both alignment sensitivity and precision relative to aligners using existing functions.

Conclusions

We developed and implemented a novel derivative-free neural network and aligner (Nepal) for optimizing sequence alignments. Nepal improved alignment quality by adapting to remote sequence alignments and increasing the expressiveness of similarity scores. Additionally, this novel scoring function can be realized using a simple matrix operation and easily incorporated into other aligners. Moreover our scoring function could potentially improve the performance of homology detection and/or multiple-sequence alignment of remote homologous sequences. The goal of the study was to provide a novel scoring function for profile alignment method and develop a novel learning system capable of addressing derivative-free problems. Our system is capable of optimizing the performance of other sophisticated methods and solving problems without derivative-of-cost functions, which do not always exist in practical problems. Our results demonstrated the usefulness of this optimization method for derivative-free problems.