Reconstructing the Genomic Content of Microbiome Taxa through Shotgun Metagenomic Deconvolution

Metagenomics has transformed our understanding of the microbial world, allowing researchers to bypass the need to isolate and culture individual taxa and to directly characterize both the taxonomic and gene compositions of environmental samples. However, associating the genes found in a metagenomic sample with the specific taxa of origin remains a critical challenge. Existing binning methods, based on nucleotide composition or alignment to reference genomes allow only a coarse-grained classification and rely heavily on the availability of sequenced genomes from closely related taxa. Here, we introduce a novel computational framework, integrating variation in gene abundances across multiple samples with taxonomic abundance data to deconvolve metagenomic samples into taxa-specific gene profiles and to reconstruct the genomic content of community members. This assembly-free method is not bounded by various factors limiting previously described methods of metagenomic binning or metagenomic assembly and represents a fundamentally different approach to metagenomic-based genome reconstruction. An implementation of this framework is available at http://elbo.gs.washington.edu/software.html. We first describe the mathematical foundations of our framework and discuss considerations for implementing its various components. We demonstrate the ability of this framework to accurately deconvolve a set of metagenomic samples and to recover the gene content of individual taxa using synthetic metagenomic samples. We specifically characterize determinants of prediction accuracy and examine the impact of annotation errors on the reconstructed genomes. We finally apply metagenomic deconvolution to samples from the Human Microbiome Project, successfully reconstructing genus-level genomic content of various microbial genera, based solely on variation in gene count. These reconstructed genera are shown to correctly capture genus-specific properties. With the accumulation of metagenomic data, this deconvolution framework provides an essential tool for characterizing microbial taxa never before seen, laying the foundation for addressing fundamental questions concerning the taxa comprising diverse microbial communities.     

Estimating the total genome length of a metagenomic sample using k-mers

Background

Metagenomic sequencing is a powerful technology for studying the mixture of microbes or the microbiomes on human and in the environment. One basic task of analyzing metagenomic data is to identify the component genomes in the community. This task is challenging due to the complexity of microbiome composition, limited availability of known reference genomes, and usually insufficient sequencing coverage.

Results

As an initial step toward understanding the complete composition of a metagenomic sample, we studied the problem of estimating the total length of all distinct component genomes in a metagenomic sample. We showed that this problem can be solved by estimating the total number of distinct k-mers in all the metagenomic sequencing data. We proposed a method for this estimation based on the sequencing coverage distribution of observed k-mers, and introduced a k-mer redundancy index (KRI) to fill in the gap between the count of distinct k-mers and the total genome length. We showed the effectiveness of the proposed method on a set of carefully designed simulation data corresponding to multiple situations of true metagenomic data. Results on real data indicate that the uncaptured genomic information can vary dramatically across metagenomic samples, with the potential to mislead downstream analyses.

Conclusions

We proposed the question of how long the total genome length of all different species in a microbial community is and introduced a method to answer it.

     

Unsupervised statistical clustering of environmental shotgun sequences

Background  

The development of effective environmental shotgun sequence binning methods remains an ongoing challenge in algorithmic analysis of metagenomic data. While previous methods have focused primarily on supervised learning involving extrinsic data, a first-principles statistical model combined with a self-training fitting method has not yet been developed.     

MetaVelvet-SL: an extension of the Velvet assembler to a de novo metagenomic assembler utilizing supervised learning

The assembly of multiple genomes from mixed sequence reads is a bottleneck in metagenomic analysis. A single-genome assembly program (assembler) is not capable of resolving metagenome sequences, so assemblers designed specifically for metagenomics have been developed. MetaVelvet is an extension of the single-genome assembler Velvet. It has been proved to generate assemblies with higher N50 scores and higher quality than single-genome assemblers such as Velvet and SOAPdenovo when applied to metagenomic sequence reads and is frequently used in this research community. One important open problem for MetaVelvet is its low accuracy and sensitivity in detecting chimeric nodes in the assembly (de Bruijn) graph, which prevents the generation of longer contigs and scaffolds. We have tackled this problem of classifying chimeric nodes using supervised machine learning to significantly improve the performance of MetaVelvet and developed a new tool, called MetaVelvet-SL. A Support Vector Machine is used for learning the classification model based on 94 features extracted from candidate nodes. In extensive experiments, MetaVelvet-SL outperformed the original MetaVelvet and other state-of-the-art metagenomic assemblers, IDBA-UD, Ray Meta and Omega, to reconstruct accurate longer assemblies with higher N50 scores for both simulated data sets and real data sets of human gut microbial sequences.     

No Longer Confidential: Estimating the Confidence of Individual Regression Predictions

Quantitative predictions in computational life sciences are often based on regression models. The advent of machine learning has led to highly accurate regression models that have gained widespread acceptance. While there are statistical methods available to estimate the global performance of regression models on a test or training dataset, it is often not clear how well this performance transfers to other datasets or how reliable an individual prediction is–a fact that often reduces a user’s trust into a computational method. In analogy to the concept of an experimental error, we sketch how estimators for individual prediction errors can be used to provide confidence intervals for individual predictions. Two novel statistical methods, named CONFINE and CONFIVE, can estimate the reliability of an individual prediction based on the local properties of nearby training data. The methods can be applied equally to linear and non-linear regression methods with very little computational overhead. We compare our confidence estimators with other existing confidence and applicability domain estimators on two biologically relevant problems (MHC–peptide binding prediction and quantitative structure-activity relationship (QSAR)). Our results suggest that the proposed confidence estimators perform comparable to or better than previously proposed estimation methods. Given a sufficient amount of training data, the estimators exhibit error estimates of high quality. In addition, we observed that the quality of estimated confidence intervals is predictable. We discuss how confidence estimation is influenced by noise, the number of features, and the dataset size. Estimating the confidence in individual prediction in terms of error intervals represents an important step from plain, non-informative predictions towards transparent and interpretable predictions that will help to improve the acceptance of computational methods in the biological community.     

基于机器学习的肠道菌群数据建模与分析研究综述

人体肠道菌群与人类的健康和疾病存在密切关系,对肠道菌群的宏基因组数据进行建模和分析,在疾病预测及诊断相关领域科学研究和社会应用方面均具有重要意义。本文从大数据分析和机器学习的角度,对人体肠道菌群数据的建模、分析和预测算法的原理、过程以及典型研究应用实例进行综述,以期推动肠道菌群分析相关研究发展以及探索结合机器学习算法进行肠道菌群分析的有效方式,同时也为开发基于肠道菌群数据的新型诊疗手段提供借鉴,推动我国精准医疗事业发展。     

A Bioinformatician's Guide to Metagenomics

Summary: As random shotgun metagenomic projects proliferate and become the dominant source of publicly available sequence data, procedures for the best practices in their execution and analysis become increasingly important. Based on our experience at the Joint Genome Institute, we describe the chain of decisions accompanying a metagenomic project from the viewpoint of the bioinformatic analysis step by step. We guide the reader through a standard workflow for a metagenomic project beginning with presequencing considerations such as community composition and sequence data type that will greatly influence downstream analyses. We proceed with recommendations for sampling and data generation including sample and metadata collection, community profiling, construction of shotgun libraries, and sequencing strategies. We then discuss the application of generic sequence processing steps (read preprocessing, assembly, and gene prediction and annotation) to metagenomic data sets in contrast to genome projects. Different types of data analyses particular to metagenomes are then presented, including binning, dominant population analysis, and gene-centric analysis. Finally, data management issues are presented and discussed. We hope that this review will assist bioinformaticians and biologists in making better-informed decisions on their journey during a metagenomic project.     

Statistical Approach of Functional Profiling for a Microbial Community

Background

Metagenomics is a relatively new but fast growing field within environmental biology and medical sciences. It enables researchers to understand the diversity of microbes, their functions, cooperation, and evolution in a particular ecosystem. Traditional methods in genomics and microbiology are not efficient in capturing the structure of the microbial community in an environment. Nowadays, high-throughput next-generation sequencing technologies are powerfully driving the metagenomic studies. However, there is an urgent need to develop efficient statistical methods and computational algorithms to rapidly analyze the massive metagenomic short sequencing data and to accurately detect the features/functions present in the microbial community. Although several issues about functions of metagenomes at pathways or subsystems level have been investigated, there is a lack of studies focusing on functional analysis at a low level of a hierarchical functional tree, such as SEED subsystem tree.

Results

A two-step statistical procedure (metaFunction) is proposed to detect all possible functional roles at the low level from a metagenomic sample/community. In the first step a statistical mixture model is proposed at the base of gene codons to estimate the abundances for the candidate functional roles, with sequencing error being considered. As a gene could be involved in multiple biological processes the functional assignment is therefore adjusted by utilizing an error distribution in the second step. The performance of the proposed procedure is evaluated through comprehensive simulation studies. Compared with other existing methods in metagenomic functional analysis the new approach is more accurate in assigning reads to functional roles, and therefore at more general levels. The method is also employed to analyze two real data sets.

Conclusions

metaFunction is a powerful tool in accurate profiling functions in a metagenomic sample.     

Efficient Prediction of Co-Complexed Proteins Based on Coevolution

The prediction of the network of protein-protein interactions (PPI) of an organism is crucial for the understanding of biological processes and for the development of new drugs. Machine learning methods have been successfully applied to the prediction of PPI in yeast by the integration of multiple direct and indirect biological data sources. However, experimental data are not available for most organisms. We propose here an ensemble machine learning approach for the prediction of PPI that depends solely on features independent from experimental data. We developed new estimators of the coevolution between proteins and combined them in an ensemble learning procedure.We applied this method to a dataset of known co-complexed proteins in Escherichia coli and compared it to previously published methods. We show that our method allows prediction of PPI with an unprecedented precision of 95.5% for the first 200 sorted pairs of proteins compared to 28.5% on the same dataset with the previous best method.A close inspection of the best predicted pairs allowed us to detect new or recently discovered interactions between chemotactic components, the flagellar apparatus and RNA polymerase complexes in E. coli.     

Predicting protein structural class based on multi-features fusion

Structural class characterizes the overall folding type of a protein or its domain and the prediction of protein structural class has become both an important and a challenging topic in protein science. Moreover, the prediction itself can stimulate the development of novel predictors that may be straightforwardly applied to many other relational areas. In this paper, 10 frequently used sequence-derived structural and physicochemical features, which can be easily computed by the PROFEAT (Protein Features) web server, were taken as inputs of support vector machines to develop statistical learning models for predicting the protein structural class. More importantly, a strategy of merging different features, called best-first search, was developed. It was shown through the rigorous jackknife cross-validation test that the success rates by our method were significantly improved. We anticipate that the present method may also have important impacts on boosting the predictive accuracies for a series of other protein attributes, such as subcellular localization, membrane types, enzyme family and subfamily classes, among many others.     

Statistical Object Data Analysis of Taxonomic Trees from Human Microbiome Data

Human microbiome research characterizes the microbial content of samples from human habitats to learn how interactions between bacteria and their host might impact human health. In this work a novel parametric statistical inference method based on object-oriented data analysis (OODA) for analyzing HMP data is proposed. OODA is an emerging area of statistical inference where the goal is to apply statistical methods to objects such as functions, images, and graphs or trees. The data objects that pertain to this work are taxonomic trees of bacteria built from analysis of 16S rRNA gene sequences (e.g. using RDP); there is one such object for each biological sample analyzed. Our goal is to model and formally compare a set of trees. The contribution of our work is threefold: first, a weighted tree structure to analyze RDP data is introduced; second, using a probability measure to model a set of taxonomic trees, we introduce an approximate MLE procedure for estimating model parameters and we derive LRT statistics for comparing the distributions of two metagenomic populations; and third the Jumpstart HMP data is analyzed using the proposed model providing novel insights and future directions of analysis.     

Microbiome-based disease prediction with multimodal variational information bottlenecks

Scientific research is shedding light on the interaction of the gut microbiome with the human host and on its role in human health. Existing machine learning methods have shown great potential in discriminating healthy from diseased microbiome states. Most of them leverage shotgun metagenomic sequencing to extract gut microbial species-relative abundances or strain-level markers. Each of these gut microbial profiling modalities showed diagnostic potential when tested separately; however, no existing approach combines them in a single predictive framework. Here, we propose the Multimodal Variational Information Bottleneck (MVIB), a novel deep learning model capable of learning a joint representation of multiple heterogeneous data modalities. MVIB achieves competitive classification performance while being faster than existing methods. Additionally, MVIB offers interpretable results. Our model adopts an information theoretic interpretation of deep neural networks and computes a joint stochastic encoding of different input data modalities. We use MVIB to predict whether human hosts are affected by a certain disease by jointly analysing gut microbial species-relative abundances and strain-level markers. MVIB is evaluated on human gut metagenomic samples from 11 publicly available disease cohorts covering 6 different diseases. We achieve high performance (0.80 < ROC AUC < 0.95) on 5 cohorts and at least medium performance on the remaining ones. We adopt a saliency technique to interpret the output of MVIB and identify the most relevant microbial species and strain-level markers to the model’s predictions. We also perform cross-study generalisation experiments, where we train and test MVIB on different cohorts of the same disease, and overall we achieve comparable results to the baseline approach, i.e. the Random Forest. Further, we evaluate our model by adding metabolomic data derived from mass spectrometry as a third input modality. Our method is scalable with respect to input data modalities and has an average training time of < 1.4 seconds. The source code and the datasets used in this work are publicly available.     

FrankSum: new feature selection method for protein function prediction

In the study of in silico functional genomics, improving the performance of protein function prediction is the ultimate goal for identifying proteins associated with defined cellular functions. The classical prediction approach is to employ pairwise sequence alignments. However this method often faces difficulties when no statistically significant homologous sequences are identified. An alternative way is to predict protein function from sequence-derived features using machine learning. In this case the choice of possible features which can be derived from the sequence is of vital importance to ensure adequate discrimination to predict function. In this paper we have successfully selected biologically significant features for protein function prediction. This was performed using a new feature selection method (FrankSum) that avoids data distribution assumptions, uses a data independent measurement (p-value) within the feature, identifies redundancy between features and uses an appropriate ranking criterion for feature selection. We have shown that classifiers generated from features selected by FrankSum outperforms classifiers generated from full feature sets, randomly selected features and features selected from the Wrapper method. We have also shown the features are concordant across all species and top ranking features are biologically informative. We conclude that feature selection is vital for successful protein function prediction and FrankSum is one of the feature selection methods that can be applied successfully to such a domain.     

Screening and purification for novel cytochrome b5 from uncultured environmental micro-organisms

AIMS: We describe a sequence-based PCR method suitable for the isolation of a novel soluble heme-binding domain of cytochrome b(5) (cyt b(5)) gene directly from metagenomic DNA is described. METHODS AND RESULTS: Using the degenerate primer set, a cyt b(5) gene was isolated directly from metagenomic DNA. Based on the sequence-based PCR method, the similar conserved motif of cyt b(5) from Rhodopseudomonas palustris strain makes the novel target gene. The gene encoding cyt b(5) was cloned and expressed in Escherichia coli BL21 (DE3) using pET expression system. The expressed recombinant enzyme was purified by Ni-nitrilotriacetic acid affinity chromatography and characterized. CONCLUSIONS: Sequence-based strategy is an effective method for application of the novel gene from metagenomic DNA. SIGNIFICANCE AND IMPACT OF THE STUDY: Investigation of novel genes from metagenome, most of the micro-organism species are largely untapped, could represent an interesting and useful reservoir for biological processes.     

CONTRAfold: RNA secondary structure prediction without physics-based models

MOTIVATION: For several decades, free energy minimization methods have been the dominant strategy for single sequence RNA secondary structure prediction. More recently, stochastic context-free grammars (SCFGs) have emerged as an alternative probabilistic methodology for modeling RNA structure. Unlike physics-based methods, which rely on thousands of experimentally-measured thermodynamic parameters, SCFGs use fully-automated statistical learning algorithms to derive model parameters. Despite this advantage, however, probabilistic methods have not replaced free energy minimization methods as the tool of choice for secondary structure prediction, as the accuracies of the best current SCFGs have yet to match those of the best physics-based models. RESULTS: In this paper, we present CONTRAfold, a novel secondary structure prediction method based on conditional log-linear models (CLLMs), a flexible class of probabilistic models which generalize upon SCFGs by using discriminative training and feature-rich scoring. In a series of cross-validation experiments, we show that grammar-based secondary structure prediction methods formulated as CLLMs consistently outperform their SCFG analogs. Furthermore, CONTRAfold, a CLLM incorporating most of the features found in typical thermodynamic models, achieves the highest single sequence prediction accuracies to date, outperforming currently available probabilistic and physics-based techniques. Our result thus closes the gap between probabilistic and thermodynamic models, demonstrating that statistical learning procedures provide an effective alternative to empirical measurement of thermodynamic parameters for RNA secondary structure prediction. AVAILABILITY: Source code for CONTRAfold is available at http://contra.stanford.edu/contrafold/.