Showing papers in "BMC Bioinformatics in 2018"

ASTRAL-III: polynomial time species tree reconstruction from partially resolved gene trees.

Abstract: Evolutionary histories can be discordant across the genome, and such discordances need to be considered in reconstructing the species phylogeny. ASTRAL is one of the leading methods for inferring species trees from gene trees while accounting for gene tree discordance. ASTRAL uses dynamic programming to search for the tree that shares the maximum number of quartet topologies with input gene trees, restricting itself to a predefined set of bipartitions. We introduce ASTRAL-III, which substantially improves the running time of ASTRAL-II and guarantees polynomial running time as a function of both the number of species (n) and the number of genes (k). ASTRAL-III limits the bipartition constraint set (X) to grow at most linearly with n and k. Moreover, it handles polytomies more efficiently than ASTRAL-II, exploits similarities between gene trees better, and uses several techniques to avoid searching parts of the search space that are mathematically guaranteed not to include the optimal tree. The asymptotic running time of ASTRAL-III in the presence of polytomies is $O\left ((nk)^{1.726} D \right)$ where D=O(nk) is the sum of degrees of all unique nodes in input trees. The running time improvements enable us to test whether contracting low support branches in gene trees improves the accuracy by reducing noise. In extensive simulations, we show that removing branches with very low support (e.g., below 10%) improves accuracy while overly aggressive filtering is harmful. We observe on a biological avian phylogenomic dataset of 14K genes that contracting low support branches greatly improve results. ASTRAL-III is a faster version of the ASTRAL method for phylogenetic reconstruction and can scale up to 10,000 species. With ASTRAL-III, low support branches can be removed, resulting in improved accuracy.

iDEP: an integrated web application for differential expression and pathway analysis of RNA-Seq data

Abstract: RNA-seq is widely used for transcriptomic profiling, but the bioinformatics analysis of resultant data can be time-consuming and challenging, especially for biologists. We aim to streamline the bioinformatic analyses of gene-level data by developing a user-friendly, interactive web application for exploratory data analysis, differential expression, and pathway analysis. iDEP (integrated Differential Expression and Pathway analysis) seamlessly connects 63 R/Bioconductor packages, 2 web services, and comprehensive annotation and pathway databases for 220 plant and animal species. The workflow can be reproduced by downloading customized R code and related pathway files. As an example, we analyzed an RNA-Seq dataset of lung fibroblasts with Hoxa1 knockdown and revealed the possible roles of SP1 and E2F1 and their target genes, including microRNAs, in blocking G1/S transition. In another example, our analysis shows that in mouse B cells without functional p53, ionizing radiation activates the MYC pathway and its downstream genes involved in cell proliferation, ribosome biogenesis, and non-coding RNA metabolism. In wildtype B cells, radiation induces p53-mediated apoptosis and DNA repair while suppressing the target genes of MYC and E2F1, and leads to growth and cell cycle arrest. iDEP helps unveil the multifaceted functions of p53 and the possible involvement of several microRNAs such as miR-92a, miR-504, and miR-30a. In both examples, we validated known molecular pathways and generated novel, testable hypotheses. Combining comprehensive analytic functionalities with massive annotation databases, iDEP ( http://ge-lab.org/idep/ ) enables biologists to easily translate transcriptomic and proteomic data into actionable insights.

Purge Haplotigs: allelic contig reassignment for third-gen diploid genome assemblies.

Abstract: Recent developments in third-gen long read sequencing and diploid-aware assemblers have resulted in the rapid release of numerous reference-quality assemblies for diploid genomes. However, assembly of highly heterozygous genomes is still problematic when regional heterogeneity is so high that haplotype homology is not recognised during assembly. This results in regional duplication rather than consolidation into allelic variants and can cause issues with downstream analysis, for example variant discovery, or haplotype reconstruction using the diploid assembly with unpaired allelic contigs. A new pipeline—Purge Haplotigs—was developed specifically for third-gen sequencing-based assemblies to automate the reassignment of allelic contigs, and to assist in the manual curation of genome assemblies. The pipeline uses a draft haplotype-fused assembly or a diploid assembly, read alignments, and repeat annotations to identify allelic variants in the primary assembly. The pipeline was tested on a simulated dataset and on four recent diploid (phased) de novo assemblies from third-generation long-read sequencing, and compared with a similar tool. After processing with Purge Haplotigs, haploid assemblies were less duplicated with minimal impact on genome completeness, and diploid assemblies had more pairings of allelic contigs. Purge Haplotigs improves the haploid and diploid representations of third-gen sequencing based genome assemblies by identifying and reassigning allelic contigs. The implementation is fast and scales well with large genomes, and it is less likely to over-purge repetitive or paralogous elements compared to alignment-only based methods. The software is available at https://bitbucket.org/mroachawri/purge_haplotigs under a permissive MIT licence.

Random forest versus logistic regression: a large-scale benchmark experiment.

Abstract: The Random Forest (RF) algorithm for regression and classification has considerably gained popularity since its introduction in 2001. Meanwhile, it has grown to a standard classification approach competing with logistic regression in many innovation-friendly scientific fields. In this context, we present a large scale benchmarking experiment based on 243 real datasets comparing the prediction performance of the original version of RF with default parameters and LR as binary classification tools. Most importantly, the design of our benchmark experiment is inspired from clinical trial methodology, thus avoiding common pitfalls and major sources of biases. RF performed better than LR according to the considered accuracy measured in approximately 69% of the datasets. The mean difference between RF and LR was 0.029 (95%-CI =[0.022,0.038]) for the accuracy, 0.041 (95%-CI =[0.031,0.053]) for the Area Under the Curve, and − 0.027 (95%-CI =[−0.034,−0.021]) for the Brier score, all measures thus suggesting a significantly better performance of RF. As a side-result of our benchmarking experiment, we observed that the results were noticeably dependent on the inclusion criteria used to select the example datasets, thus emphasizing the importance of clear statements regarding this dataset selection process. We also stress that neutral studies similar to ours, based on a high number of datasets and carefully designed, will be necessary in the future to evaluate further variants, implementations or parameters of random forests which may yield improved accuracy compared to the original version with default values.

Rapid and precise alignment of raw reads against redundant databases with KMA.

Abstract: As the cost of sequencing has declined, clinical diagnostics based on next generation sequencing (NGS) have become reality. Diagnostics based on sequencing will require rapid and precise mapping against redundant databases because some of the most important determinants, such as antimicrobial resistance and core genome multilocus sequence typing (MLST) alleles, are highly similar to one another. In order to facilitate this, a novel mapping method, KMA (k-mer alignment), was designed. KMA is able to map raw reads directly against redundant databases, it also scales well for large redundant databases. KMA uses k-mer seeding to speed up mapping and the Needleman-Wunsch algorithm to accurately align extensions from k-mer seeds. Multi-mapping reads are resolved using a novel sorting scheme (ConClave scheme), ensuring an accurate selection of templates. The functionality of KMA was compared with SRST2, MGmapper, BWA-MEM, Bowtie2, Minimap2 and Salmon, using both simulated data and a dataset of Escherichia coli mapped against resistance genes and core genome MLST alleles. KMA outperforms current methods with respect to both accuracy and speed, while using a comparable amount of memory. With KMA, it was possible map raw reads directly against redundant databases with high accuracy, speed and memory efficiency.

Recommending plant taxa for supporting on-site species identification

Abstract: Predicting a list of plant taxa most likely to be observed at a given geographical location and time is useful for many scenarios in biodiversity informatics. Since efficient plant species identification is impeded mainly by the large number of possible candidate species, providing a shortlist of likely candidates can help significantly expedite the task. Whereas species distribution models heavily rely on geo-referenced occurrence data, such information still remains largely unused for plant taxa identification tools. In this paper, we conduct a study on the feasibility of computing a ranked shortlist of plant taxa likely to be encountered by an observer in the field. We use the territory of Germany as case study with a total of 7.62M records of freely available plant presence-absence data and occurrence records for 2.7k plant taxa. We systematically study achievable recommendation quality based on two types of source data: binary presence-absence data and individual occurrence records. Furthermore, we study strategies for aggregating records into a taxa recommendation based on location and date of an observation. We evaluate recommendations using 28k geo-referenced and taxa-labeled plant images hosted on the Flickr website as an independent test dataset. Relying on location information from presence-absence data alone results in an average recall of 82%. However, we find that occurrence records are complementary to presence-absence data and using both in combination yields considerably higher recall of 96% along with improved ranking metrics. Ultimately, by reducing the list of candidate taxa by an average of 62%, a spatio-temporal prior can substantially expedite the overall identification problem.

DrImpute: imputing dropout events in single cell RNA sequencing data.

Abstract: The single cell RNA sequencing (scRNA-seq) technique begin a new era by allowing the observation of gene expression at the single cell level. However, there is also a large amount of technical and biological noise. Because of the low number of RNA transcriptomes and the stochastic nature of the gene expression pattern, there is a high chance of missing nonzero entries as zero, which are called dropout events. We develop DrImpute to impute dropout events in scRNA-seq data. We show that DrImpute has significantly better performance on the separation of the dropout zeros from true zeros than existing imputation algorithms. We also demonstrate that DrImpute can significantly improve the performance of existing tools for clustering, visualization and lineage reconstruction of nine published scRNA-seq datasets. DrImpute can serve as a very useful addition to the currently existing statistical tools for single cell RNA-seq analysis. DrImpute is implemented in R and is available at https://github.com/gongx030/DrImpute .

Single sample scoring of molecular phenotypes

Abstract: Gene set scoring provides a useful approach for quantifying concordance between sample transcriptomes and selected molecular signatures. Most methods use information from all samples to score an individual sample, leading to unstable scores in small data sets and introducing biases from sample composition (e.g. varying numbers of samples for different cancer subtypes). To address these issues, we have developed a truly single sample scoring method, and associated R/Bioconductor package singscore ( https://bioconductor.org/packages/singscore ). We use multiple cancer data sets to compare singscore against widely-used methods, including GSVA, z-score, PLAGE, and ssGSEA. Our approach does not depend upon background samples and scores are thus stable regardless of the composition and number of samples being scored. In contrast, scores obtained by GSVA, z-score, PLAGE and ssGSEA can be unstable when less data are available (NS < 25). The singscore method performs as well as the best performing methods in terms of power, recall, false positive rate and computational time, and provides consistently high and balanced performance across all these criteria. To enhance the impact and utility of our method, we have also included a set of functions implementing visual analysis and diagnostics to support the exploration of molecular phenotypes in single samples and across populations of data. The singscore method described here functions independent of sample composition in gene expression data and thus it provides stable scores, which are particularly useful for small data sets or data integration. Singscore performs well across all performance criteria, and includes a suite of powerful visualization functions to assist in the interpretation of results. This method performs as well as or better than other scoring approaches in terms of its power to distinguish samples with distinct biology and its ability to call true differential gene sets between two conditions. These scores can be used for dimensional reduction of transcriptomic data and the phenotypic landscapes obtained by scoring samples against multiple molecular signatures may provide insights for sample stratification.

SVM-RFE: selection and visualization of the most relevant features through non-linear kernels

Abstract: Support vector machines (SVM) are a powerful tool to analyze data with a number of predictors approximately equal or larger than the number of observations. However, originally, application of SVM to analyze biomedical data was limited because SVM was not designed to evaluate importance of predictor variables. Creating predictor models based on only the most relevant variables is essential in biomedical research. Currently, substantial work has been done to allow assessment of variable importance in SVM models but this work has focused on SVM implemented with linear kernels. The power of SVM as a prediction model is associated with the flexibility generated by use of non-linear kernels. Moreover, SVM has been extended to model survival outcomes. This paper extends the Recursive Feature Elimination (RFE) algorithm by proposing three approaches to rank variables based on non-linear SVM and SVM for survival analysis. The proposed algorithms allows visualization of each one the RFE iterations, and hence, identification of the most relevant predictors of the response variable. Using simulation studies based on time-to-event outcomes and three real datasets, we evaluate the three methods, based on pseudo-samples and kernel principal component analysis, and compare them with the original SVM-RFE algorithm for non-linear kernels. The three algorithms we proposed performed generally better than the gold standard RFE for non-linear kernels, when comparing the truly most relevant variables with the variable ranks produced by each algorithm in simulation studies. Generally, the RFE-pseudo-samples outperformed the other three methods, even when variables were assumed to be correlated in all tested scenarios. The proposed approaches can be implemented with accuracy to select variables and assess direction and strength of associations in analysis of biomedical data using SVM for categorical or time-to-event responses. Conducting variable selection and interpreting direction and strength of associations between predictors and outcomes with the proposed approaches, particularly with the RFE-pseudo-samples approach can be implemented with accuracy when analyzing biomedical data. These approaches, perform better than the classical RFE of Guyon for realistic scenarios about the structure of biomedical data.

Evaluating methods of inferring gene regulatory networks highlights their lack of performance for single cell gene expression data

Abstract: A fundamental fact in biology states that genes do not operate in isolation, and yet, methods that infer regulatory networks for single cell gene expression data have been slow to emerge. With single cell sequencing methods now becoming accessible, general network inference algorithms that were initially developed for data collected from bulk samples may not be suitable for single cells. Meanwhile, although methods that are specific for single cell data are now emerging, whether they have improved performance over general methods is unknown. In this study, we evaluate the applicability of five general methods and three single cell methods for inferring gene regulatory networks from both experimental single cell gene expression data and in silico simulated data. Standard evaluation metrics using ROC curves and Precision-Recall curves against reference sets sourced from the literature demonstrated that most of the methods performed poorly when they were applied to either experimental single cell data, or simulated single cell data, which demonstrates their lack of performance for this task. Using default settings, network methods were applied to the same datasets. Comparisons of the learned networks highlighted the uniqueness of some predicted edges for each method. The fact that different methods infer networks that vary substantially reflects the underlying mathematical rationale and assumptions that distinguish network methods from each other. This study provides a comprehensive evaluation of network modeling algorithms applied to experimental single cell gene expression data and in silico simulated datasets where the network structure is known. Comparisons demonstrate that most of these assessed network methods are not able to predict network structures from single cell expression data accurately, even if they are specifically developed for single cell methods. Also, single cell methods, which usually depend on more elaborative algorithms, in general have less similarity to each other in the sets of edges detected. The results from this study emphasize the importance for developing more accurate optimized network modeling methods that are compatible for single cell data. Newly-developed single cell methods may uniquely capture particular features of potential gene-gene relationships, and caution should be taken when we interpret these results.

CEMiTool: a Bioconductor package for performing comprehensive modular co-expression analyses.

Abstract: The analysis of modular gene co-expression networks is a well-established method commonly used for discovering the systems-level functionality of genes. In addition, these studies provide a basis for the discovery of clinically relevant molecular pathways underlying different diseases and conditions. In this paper, we present a fast and easy-to-use Bioconductor package named CEMiTool that unifies the discovery and the analysis of co-expression modules. Using the same real datasets, we demonstrate that CEMiTool outperforms existing tools, and provides unique results in a user-friendly html report with high quality graphs. Among its features, our tool evaluates whether modules contain genes that are over-represented by specific pathways or that are altered in a specific sample group, as well as it integrates transcriptomic data with interactome information, identifying the potential hubs on each network. We successfully applied CEMiTool to over 1000 transcriptome datasets, and to a new RNA-seq dataset of patients infected with Leishmania, revealing novel insights of the disease’s physiopathology. The CEMiTool R package provides users with an easy-to-use method to automatically implement gene co-expression network analyses, obtain key information about the discovered gene modules using additional downstream analyses and retrieve publication-ready results via a high-quality interactive report.

Evaluation and comparison of bioinformatic tools for the enrichment analysis of metabolomics data

Abstract: Bioinformatic tools for the enrichment of ‘omics’ datasets facilitate interpretation and understanding of data. To date few are suitable for metabolomics datasets. The main objective of this work is to give a critical overview, for the first time, of the performance of these tools. To that aim, datasets from metabolomic repositories were selected and enriched data were created. Both types of data were analysed with these tools and outputs were thoroughly examined. An exploratory multivariate analysis of the most used tools for the enrichment of metabolite sets, based on a non-metric multidimensional scaling (NMDS) of Jaccard’s distances, was performed and mirrored their diversity. Codes (identifiers) of the metabolites of the datasets were searched in different metabolite databases (HMDB, KEGG, PubChem, ChEBI, BioCyc/HumanCyc, LipidMAPS, ChemSpider, METLIN and Recon2). The databases that presented more identifiers of the metabolites of the dataset were PubChem, followed by METLIN and ChEBI. However, these databases had duplicated entries and might present false positives. The performance of over-representation analysis (ORA) tools, including BioCyc/HumanCyc, ConsensusPathDB, IMPaLA, MBRole, MetaboAnalyst, Metabox, MetExplore, MPEA, PathVisio and Reactome and the mapping tool KEGGREST, was examined. Results were mostly consistent among tools and between real and enriched data despite the variability of the tools. Nevertheless, a few controversial results such as differences in the total number of metabolites were also found. Disease-based enrichment analyses were also assessed, but they were not found to be accurate probably due to the fact that metabolite disease sets are not up-to-date and the difficulty of predicting diseases from a list of metabolites. We have extensively reviewed the state-of-the-art of the available range of tools for metabolomic datasets, the completeness of metabolite databases, the performance of ORA methods and disease-based analyses. Despite the variability of the tools, they provided consistent results independent of their analytic approach. However, more work on the completeness of metabolite and pathway databases is required, which strongly affects the accuracy of enrichment analyses. Improvements will be translated into more accurate and global insights of the metabolome.

GOnet: a tool for interactive Gene Ontology analysis.

Abstract: Biological interpretation of gene/protein lists resulting from -omics experiments can be a complex task. A common approach consists of reviewing Gene Ontology (GO) annotations for entries in such lists and searching for enrichment patterns. Unfortunately, there is a gap between machine-readable output of GO software and its human-interpretable form. This gap can be bridged by allowing users to simultaneously visualize and interact with term-term and gene-term relationships. We created the open-source GOnet web-application (available at http://tools.dice-database.org/GOnet/ ), which takes a list of gene or protein entries from human or mouse data and performs GO term annotation analysis (mapping of provided entries to GO subsets) or GO term enrichment analysis (scanning for GO categories overrepresented in the input list). The application is capable of producing parsable data formats and importantly, interactive visualizations of the GO analysis results. The interactive results allow exploration of genes and GO terms as a graph that depicts the natural hierarchy of the terms and retains relationships between terms and genes/proteins. As a result, GOnet provides insight into the functional interconnection of the submitted entries. The application can be used for GO analysis of any biological data sources resulting in gene/protein lists. It can be helpful for experimentalists as well as computational biologists working on biological interpretation of -omics data resulting in such lists.

Predicting drug-disease associations by using similarity constrained matrix factorization

Abstract: Drug-disease associations provide important information for the drug discovery. Wet experiments that identify drug-disease associations are time-consuming and expensive. However, many drug-disease associations are still unobserved or unknown. The development of computational methods for predicting unobserved drug-disease associations is an important and urgent task. In this paper, we proposed a similarity constrained matrix factorization method for the drug-disease association prediction (SCMFDD), which makes use of known drug-disease associations, drug features and disease semantic information. SCMFDD projects the drug-disease association relationship into two low-rank spaces, which uncover latent features for drugs and diseases, and then introduces drug feature-based similarities and disease semantic similarity as constraints for drugs and diseases in low-rank spaces. Different from the classic matrix factorization technique, SCMFDD takes the biological context of the problem into account. In computational experiments, the proposed method can produce high-accuracy performances on benchmark datasets, and outperform existing state-of-the-art prediction methods when evaluated by five-fold cross validation and independent testing. We developed a user-friendly web server by using known associations collected from the CTD database, available at http://www.bioinfotech.cn/SCMFDD/ . The case studies show that the server can find out novel associations, which are not included in the CTD database.

A deep learning-based algorithm for 2-D cell segmentation in microscopy images

Abstract: Automatic and reliable characterization of cells in cell cultures is key to several applications such as cancer research and drug discovery. Given the recent advances in light microscopy and the need for accurate and high-throughput analysis of cells, automated algorithms have been developed for segmenting and analyzing the cells in microscopy images. Nevertheless, accurate, generic and robust whole-cell segmentation is still a persisting need to precisely quantify its morphological properties, phenotypes and sub-cellular dynamics. We present a single-channel whole cell segmentation algorithm. We use markers that stain the whole cell, but with less staining in the nucleus, and without using a separate nuclear stain. We show the utility of our approach in microscopy images of cell cultures in a wide variety of conditions. Our algorithm uses a deep learning approach to learn and predict locations of the cells and their nuclei, and combines that with thresholding and watershed-based segmentation. We trained and validated our approach using different sets of images, containing cells stained with various markers and imaged at different magnifications. Our approach achieved a 86% similarity to ground truth segmentation when identifying and separating cells. The proposed algorithm is able to automatically segment cells from single channel images using a variety of markers and magnifications.

Combining RNA-seq data and homology-based gene prediction for plants, animals and fungi

Abstract: Genome annotation is of key importance in many research questions. The identification of protein-coding genes is often based on transcriptome sequencing data, ab-initio or homology-based prediction. Recently, it was demonstrated that intron position conservation improves homology-based gene prediction, and that experimental data improves ab-initio gene prediction. Here, we present an extension of the gene prediction program GeMoMa that utilizes amino acid sequence conservation, intron position conservation and optionally RNA-seq data for homology-based gene prediction. We show on published benchmark data for plants, animals and fungi that GeMoMa performs better than the gene prediction programs BRAKER1, MAKER2, and CodingQuarry, and purely RNA-seq-based pipelines for transcript identification. In addition, we demonstrate that using multiple reference organisms may help to further improve the performance of GeMoMa. Finally, we apply GeMoMa to four nematode species and to the recently published barley reference genome indicating that current annotations of protein-coding genes may be refined using GeMoMa predictions. GeMoMa might be of great utility for annotating newly sequenced genomes but also for finding homologs of a specific gene or gene family. GeMoMa has been published under GNU GPL3 and is freely available at http://www.jstacs.de/index.php/GeMoMa .

VIPER: Visualization Pipeline for RNA-seq, a Snakemake workflow for efficient and complete RNA-seq analysis

Abstract: RNA sequencing has become a ubiquitous technology used throughout life sciences as an effective method of measuring RNA abundance quantitatively in tissues and cells. The increase in use of RNA-seq technology has led to the continuous development of new tools for every step of analysis from alignment to downstream pathway analysis. However, effectively using these analysis tools in a scalable and reproducible way can be challenging, especially for non-experts. Using the workflow management system Snakemake we have developed a user friendly, fast, efficient, and comprehensive pipeline for RNA-seq analysis. VIPER (Visualization Pipeline for RNA-seq analysis) is an analysis workflow that combines some of the most popular tools to take RNA-seq analysis from raw sequencing data, through alignment and quality control, into downstream differential expression and pathway analysis. VIPER has been created in a modular fashion to allow for the rapid incorporation of new tools to expand the capabilities. This capacity has already been exploited to include very recently developed tools that explore immune infiltrate and T-cell CDR (Complementarity-Determining Regions) reconstruction abilities. The pipeline has been conveniently packaged such that minimal computational skills are required to download and install the dozens of software packages that VIPER uses. VIPER is a comprehensive solution that performs most standard RNA-seq analyses quickly and effectively with a built-in capacity for customization and expansion.

Assessing microscope image focus quality with deep learning.

Abstract: Large image datasets acquired on automated microscopes typically have some fraction of low quality, out-of-focus images, despite the use of hardware autofocus systems. Identification of these images using automated image analysis with high accuracy is important for obtaining a clean, unbiased image dataset. Complicating this task is the fact that image focus quality is only well-defined in foreground regions of images, and as a result, most previous approaches only enable a computation of the relative difference in quality between two or more images, rather than an absolute measure of quality. We present a deep neural network model capable of predicting an absolute measure of image focus on a single image in isolation, without any user-specified parameters. The model operates at the image-patch level, and also outputs a measure of prediction certainty, enabling interpretable predictions. The model was trained on only 384 in-focus Hoechst (nuclei) stain images of U2OS cells, which were synthetically defocused to one of 11 absolute defocus levels during training. The trained model can generalize on previously unseen real Hoechst stain images, identifying the absolute image focus to within one defocus level (approximately 3 pixel blur diameter difference) with 95% accuracy. On a simpler binary in/out-of-focus classification task, the trained model outperforms previous approaches on both Hoechst and Phalloidin (actin) stain images (F-scores of 0.89 and 0.86, respectively over 0.84 and 0.83), despite only having been presented Hoechst stain images during training. Lastly, we observe qualitatively that the model generalizes to two additional stains, Hoechst and Tubulin, of an unseen cell type (Human MCF-7) acquired on a different instrument. Our deep neural network enables classification of out-of-focus microscope images with both higher accuracy and greater precision than previous approaches via interpretable patch-level focus and certainty predictions. The use of synthetically defocused images precludes the need for a manually annotated training dataset. The model also generalizes to different image and cell types. The framework for model training and image prediction is available as a free software library and the pre-trained model is available for immediate use in Fiji (ImageJ) and CellProfiler.

Web-based design and analysis tools for CRISPR base editing

Abstract: As a result of its simplicity and high efficiency, the CRISPR-Cas system has been widely used as a genome editing tool Recently, CRISPR base editors, which consist of deactivated Cas9 (dCas9) or Cas9 nickase (nCas9) linked with a cytidine or a guanine deaminase, have been developed Base editing tools will be very useful for gene correction because they can produce highly specific DNA substitutions without the introduction of any donor DNA, but dedicated web-based tools to facilitate the use of such tools have not yet been developed We present two web tools for base editors, named BE-Designer and BE-Analyzer BE-Designer provides all possible base editor target sequences in a given input DNA sequence with useful information including potential off-target sites BE-Analyzer, a tool for assessing base editing outcomes from next generation sequencing (NGS) data, provides information about mutations in a table and interactive graphs Furthermore, because the tool runs client-side, large amounts of targeted deep sequencing data (< 1 GB) do not need to be uploaded to a server, substantially reducing running time and increasing data security BE-Designer and BE-Analyzer can be freely accessed at http://wwwrgenomenet/be-designer/ and http://wwwrgenomenet/be-analyzer /, respectively We develop two useful web tools to design target sequence (BE-Designer) and to analyze NGS data from experimental results (BE-Analyzer) for CRISPR base editors

NGmerge: merging paired-end reads via novel empirically-derived models of sequencing errors

Abstract: Advances in Illumina DNA sequencing technology have produced longer paired-end reads that increasingly have sequence overlaps. These reads can be merged into a single read that spans the full length of the original DNA fragment, allowing for error correction and accurate determination of read coverage. Extant merging programs utilize simplistic or unverified models for the selection of bases and quality scores for the overlapping region of merged reads. We first examined the baseline quality score - error rate relationship using sequence reads derived from PhiX. In contrast to numerous published reports, we found that the quality scores produced by Illumina were not substantially inflated above the theoretical values, once the reference genome was corrected for unreported sequence variants. The PhiX reads were then used to create empirical models of sequencing errors in overlapping regions of paired-end reads, and these models were incorporated into a novel merging program, NGmerge. We demonstrate that NGmerge corrects errors and ambiguous bases better than other merging programs, and that it assigns quality scores for merged bases that accurately reflect the error rates. Our results also show that, contrary to published analyses, the sequencing errors of paired-end reads are not independent. We provide a free and open-source program, NGmerge, that performs better than existing read merging programs. NGmerge is available on GitHub ( https://github.com/harvardinformatics/NGmerge ) under the MIT License; it is written in C and supported on Linux.

lme4qtl: linear mixed models with flexible covariance structure for genetic studies of related individuals

Abstract: Quantitative trait locus (QTL) mapping in genetic data often involves analysis of correlated observations, which need to be accounted for to avoid false association signals. This is commonly performed by modeling such correlations as random effects in linear mixed models (LMMs). The R package lme4 is a well-established tool that implements major LMM features using sparse matrix methods; however, it is not fully adapted for QTL mapping association and linkage studies. In particular, two LMM features are lacking in the base version of lme4: the definition of random effects by custom covariance matrices; and parameter constraints, which are essential in advanced QTL models. Apart from applications in linkage studies of related individuals, such functionalities are of high interest for association studies in situations where multiple covariance matrices need to be modeled, a scenario not covered by many genome-wide association study (GWAS) software. To address the aforementioned limitations, we developed a new R package lme4qtl as an extension of lme4. First, lme4qtl contributes new models for genetic studies within a single tool integrated with lme4 and its companion packages. Second, lme4qtl offers a flexible framework for scenarios with multiple levels of relatedness and becomes efficient when covariance matrices are sparse. We showed the value of our package using real family-based data in the Genetic Analysis of Idiopathic Thrombophilia 2 (GAIT2) project. Our software lme4qtl enables QTL mapping models with a versatile structure of random effects and efficient computation for sparse covariances. lme4qtl is available at https://github.com/variani/lme4qtl .

Convolutional neural network based on SMILES representation of compounds for detecting chemical motif.

Abstract: Previous studies have suggested deep learning to be a highly effective approach for screening lead compounds for new drugs. Several deep learning models have been developed by addressing the use of various kinds of fingerprints and graph convolution architectures. However, these methods are either advantageous or disadvantageous depending on whether they (1) can distinguish structural differences including chirality of compounds, and (2) can automatically discover effective features. We developed another deep learning model for compound classification. In this method, we constructed a distributed representation of compounds based on the SMILES notation, which linearly represents a compound structure, and applied the SMILES-based representation to a convolutional neural network (CNN). The use of SMILES allows us to process all types of compounds while incorporating a broad range of structure information, and representation learning by CNN automatically acquires a low-dimensional representation of input features. In a benchmark experiment using the TOX 21 dataset, our method outperformed conventional fingerprint methods, and performed comparably against the winning model of the TOX 21 Challenge. Multivariate analysis confirmed that the chemical space consisting of the features learned by SMILES-based representation learning adequately expressed a richer feature space that enabled the accurate discrimination of compounds. Using motif detection with the learned filters, not only important known structures (motifs) such as protein-binding sites but also structures of unknown functional groups were detected. The source code of our SMILES-based convolutional neural network software in the deep learning framework Chainer is available at http://www.dna.bio.keio.ac.jp/smiles/ , and the dataset used for performance evaluation in this work is available at the same URL.

Genome-wide prediction of cis-regulatory regions using supervised deep learning methods

Abstract: In the human genome, 98% of DNA sequences are non-protein-coding regions that were previously disregarded as junk DNA. In fact, non-coding regions host a variety of cis-regulatory regions which precisely control the expression of genes. Thus, Identifying active cis-regulatory regions in the human genome is critical for understanding gene regulation and assessing the impact of genetic variation on phenotype. The developments of high-throughput sequencing and machine learning technologies make it possible to predict cis-regulatory regions genome wide. Based on rich data resources such as the Encyclopedia of DNA Elements (ENCODE) and the Functional Annotation of the Mammalian Genome (FANTOM) projects, we introduce DECRES based on supervised deep learning approaches for the identification of enhancer and promoter regions in the human genome. Due to their ability to discover patterns in large and complex data, the introduction of deep learning methods enables a significant advance in our knowledge of the genomic locations of cis-regulatory regions. Using models for well-characterized cell lines, we identify key experimental features that contribute to the predictive performance. Applying DECRES, we delineate locations of 300,000 candidate enhancers genome wide (6.8% of the genome, of which 40,000 are supported by bidirectional transcription data), and 26,000 candidate promoters (0.6% of the genome). The predicted annotations of cis-regulatory regions will provide broad utility for genome interpretation from functional genomics to clinical applications. The DECRES model demonstrates potentials of deep learning technologies when combined with high-throughput sequencing data, and inspires the development of other advanced neural network models for further improvement of genome annotations.

Prediction of 8-state protein secondary structures by a novel deep learning architecture

Abstract: Protein secondary structure can be regarded as an information bridge that links the primary sequence and tertiary structure. Accurate 8-state secondary structure prediction can significantly give more precise and high resolution on structure-based properties analysis. We present a novel deep learning architecture which exploits an integrative synergy of prediction by a convolutional neural network, residual network, and bidirectional recurrent neural network to improve the performance of protein secondary structure prediction. A local block comprised of convolutional filters and original input is designed for capturing local sequence features. The subsequent bidirectional recurrent neural network consisting of gated recurrent units can capture global context features. Furthermore, the residual network can improve the information flow between the hidden layers and the cascaded recurrent neural network. Our proposed deep network achieved 71.4% accuracy on the benchmark CB513 dataset for the 8-state prediction; and the ensemble learning by our model achieved 74% accuracy. Our model generalization capability is also evaluated on other three independent datasets CASP10, CASP11 and CASP12 for both 8- and 3-state prediction. These prediction performances are superior to the state-of-the-art methods. Our experiment demonstrates that it is a valuable method for predicting protein secondary structure, and capturing local and global features concurrently is very useful in deep learning.

LocText: relation extraction of protein localizations to assist database curation.

Abstract: The subcellular localization of a protein is an important aspect of its function However, the experimental annotation of locations is not even complete for well-studied model organisms Text mining might aid database curators to add experimental annotations from the scientific literature Existing extraction methods have difficulties to distinguish relationships between proteins and cellular locations co-mentioned in the same sentence LocText was created as a new method to extract protein locations from abstracts and full texts LocText learned patterns from syntax parse trees and was trained and evaluated on a newly improved LocTextCorpus Combined with an automatic named-entity recognizer, LocText achieved high precision (P = 86%±4) After completing development, we mined the latest research publications for three organisms: human (Homo sapiens), budding yeast (Saccharomyces cerevisiae), and thale cress (Arabidopsis thaliana) Examining 60 novel, text-mined annotations, we found that 65% (human), 85% (yeast), and 80% (cress) were correct Of all validated annotations, 40% were completely novel, ie did neither appear in the annotations nor the text descriptions of Swiss-Prot LocText provides a cost-effective, semi-automated workflow to assist database curators in identifying novel protein localization annotations The annotations suggested through text-mining would be verified by experts to guarantee high-quality standards of manually-curated databases such as Swiss-Prot

HiCcompare: an R-package for joint normalization and comparison of HI-C datasets

Abstract: Changes in spatial chromatin interactions are now emerging as a unifying mechanism orchestrating the regulation of gene expression. Hi-C sequencing technology allows insight into chromatin interactions on a genome-wide scale. However, Hi-C data contains many DNA sequence- and technology-driven biases. These biases prevent effective comparison of chromatin interactions aimed at identifying genomic regions differentially interacting between, e.g., disease-normal states or different cell types. Several methods have been developed for normalizing individual Hi-C datasets. However, they fail to account for biases between two or more Hi-C datasets, hindering comparative analysis of chromatin interactions. We developed a simple and effective method, HiCcompare, for the joint normalization and differential analysis of multiple Hi-C datasets. The method introduces a distance-centric analysis and visualization of the differences between two Hi-C datasets on a single plot that allows for a data-driven normalization of biases using locally weighted linear regression (loess). HiCcompare outperforms methods for normalizing individual Hi-C datasets and methods for differential analysis (diffHiC, FIND) in detecting a priori known chromatin interaction differences while preserving the detection of genomic structures, such as A/B compartments. HiCcompare is able to remove between-dataset bias present in Hi-C matrices. It also provides a user-friendly tool to allow the scientific community to perform direct comparisons between the growing number of pre-processed Hi-C datasets available at online repositories. HiCcompare is freely available as a Bioconductor R package https://bioconductor.org/packages/HiCcompare/ .

FMLRC: Hybrid long read error correction using an FM-index

Abstract: Long read sequencing is changing the landscape of genomic research, especially de novo assembly. Despite the high error rate inherent to long read technologies, increased read lengths dramatically improve the continuity and accuracy of genome assemblies. However, the cost and throughput of these technologies limits their application to complex genomes. One solution is to decrease the cost and time to assemble novel genomes by leveraging “hybrid” assemblies that use long reads for scaffolding and short reads for accuracy. We describe a novel method leveraging a multi-string Burrows-Wheeler Transform with auxiliary FM-index to correct errors in long read sequences using a set of complementary short reads. We demonstrate that our method efficiently produces significantly more high quality corrected sequence than existing hybrid error-correction methods. We also show that our method produces more contiguous assemblies, in many cases, than existing state-of-the-art hybrid and long-read only de novo assembly methods. Our method accurately corrects long read sequence data using complementary short reads. We demonstrate higher total throughput of corrected long reads and a corresponding increase in contiguity of the resulting de novo assemblies. Improved throughput and computational efficiency than existing methods will help better economically utilize emerging long read sequencing technologies.

Predicting adverse drug reactions through interpretable deep learning framework.

Abstract: 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. 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. 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. 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.

ECPred: a tool for the prediction of the enzymatic functions of protein sequences based on the EC nomenclature.

Abstract: The automated prediction of the enzymatic functions of uncharacterized proteins is a crucial topic in bioinformatics. Although several methods and tools have been proposed to classify enzymes, most of these studies are limited to specific functional classes and levels of the Enzyme Commission (EC) number hierarchy. Besides, most of the previous methods incorporated only a single input feature type, which limits the applicability to the wide functional space. Here, we proposed a novel enzymatic function prediction tool, ECPred, based on ensemble of machine learning classifiers. In ECPred, each EC number constituted an individual class and therefore, had an independent learning model. Enzyme vs. non-enzyme classification is incorporated into ECPred along with a hierarchical prediction approach exploiting the tree structure of the EC nomenclature. ECPred provides predictions for 858 EC numbers in total including 6 main classes, 55 subclass classes, 163 sub-subclass classes and 634 substrate classes. The proposed method is tested and compared with the state-of-the-art enzyme function prediction tools by using independent temporal hold-out and no-Pfam datasets constructed during this study. ECPred is presented both as a stand-alone and a web based tool to provide probabilistic enzymatic function predictions (at all five levels of EC) for uncharacterized protein sequences. Also, the datasets of this study will be a valuable resource for future benchmarking studies. ECPred is available for download, together with all of the datasets used in this study, at: https://github.com/cansyl/ECPred . ECPred webserver can be accessed through http://cansyl.metu.edu.tr/ECPred.html .

DeepM6ASeq: prediction and characterization of m6A-containing sequences using deep learning

Abstract: N6-methyladensine (m6A) is a common and abundant RNA methylation modification found in various species. As a type of post-transcriptional methylation, m6A plays an important role in diverse RNA activities such as alternative splicing, an interplay with microRNAs and translation efficiency. Although existing tools can predict m6A at single-base resolution, it is still challenging to extract the biological information surrounding m6A sites. We implemented a deep learning framework, named DeepM6ASeq, to predict m6A-containing sequences and characterize surrounding biological features based on miCLIP-Seq data, which detects m6A sites at single-base resolution. DeepM6ASeq showed better performance as compared to other machine learning classifiers. Moreover, an independent test on m6A-Seq data, which identifies m6A-containing genomic regions, revealed that our model is competitive in predicting m6A-containing sequences. The learned motifs from DeepM6ASeq correspond to known m6A readers. Notably, DeepM6ASeq also identifies a newly recognized m6A reader: FMR1. Besides, we found that a saliency map in the deep learning model could be utilized to visualize locations of m6A sites. We developed a deep-learning-based framework to predict and characterize m6A-containing sequences and hope to help investigators to gain more insights for m6A research. The source code is available at https://github.com/rreybeyb/DeepM6ASeq .