RNN-VirSeeker: A Deep Learning Method for Identification of Short Viral Sequences From Metagenomes

Deep learning methods in metagenomics: a review

The ever-decreasing cost of sequencing and the growing potential applications of metagenomics have led to an unprecedented surge in data generation. One of the most prevalent applications of metagenomics is the study of microbial environments, such as the human gut. The gut microbiome plays a crucial role in human health, providing vital information for patient diagnosis and prognosis. However, analysing metagenomic data remains challenging due to several factors, including reference catalogues, sparsity and compositionality. Deep learning (DL) enables novel and promising approaches that complement state-of-the-art microbiome pipelines. DL-based methods can address almost all aspects of microbiome analysis, including novel pathogen detection, sequence classification, patient stratification and disease prediction. Beyond generating predictive models, a key aspect of these methods is also their interpretability. This article reviews DL approaches in metagenomics, including convolutional networks, autoencoders and attention-based models. These methods aggregate contextualized data and pave the way for improved patient care and a better understanding of the microbiome's key role in our health.

Benchmarking informatics approaches for virus discovery: caution is needed when combining in silico identification methods

Understanding the ecological impacts of viruses on natural and engineered ecosystems relies on the accurate identification of viral sequences from community sequencing data. To maximize viral recovery from metagenomes, researchers frequently combine viral identification tools. However, the effectiveness of this strategy is unknown. Here, we benchmarked combinations of six widely used informatics tools for viral identification and analysis (VirSorter, VirSorter2, VIBRANT, DeepVirFinder, CheckV, and Kaiju), called "rulesets." Rulesets were tested against mock metagenomes composed of taxonomically diverse sequence types and diverse aquatic metagenomes to assess the effects of the degree of viral enrichment and habitat on tool performance. We found that six rulesets achieved equivalent accuracy [Matthews Correlation Coefficient (MCC) = 0.77, Padj ≥ 0.05]. Each contained VirSorter2, and five used our "tuning removal" rule designed to remove non-viral contamination. While DeepVirFinder, VIBRANT, and VirSorter were each found once in these high-accuracy rulesets, they were not found in combination with each other: combining tools does not lead to optimal performance. Our validation suggests that the MCC plateau at 0.77 is partly caused by inaccurate labeling within reference sequence databases. In aquatic metagenomes, our highest MCC ruleset identified more viral sequences in virus-enriched (44%-46%) than in cellular metagenomes (7%-19%). While improved algorithms may lead to more accurate viral identification tools, this should be done in tandem with careful curation of sequence databases. We recommend using the VirSorter2 ruleset and our empirically derived tuning removal rule. Our analysis provides insight into methods for in silico viral identification and will enable more robust viral identification from metagenomic data sets.

IMPORTANCE

The identification of viruses from environmental metagenomes using informatics tools has offered critical insights in microbial ecology. However, it remains difficult for researchers to know which tools optimize viral recovery for their specific study. In an attempt to recover more viruses, studies are increasingly combining the outputs from multiple tools without validating this approach. After benchmarking combinations of six viral identification tools against mock metagenomes and environmental samples, we found that these tools should only be combined cautiously. Two to four tool combinations maximized viral recovery and minimized non-viral contamination compared with either the single-tool or the five- to six-tool ones. By providing a rigorous overview of the behavior of in silico viral identification strategies and a pipeline to replicate our process, our findings guide the use of existing viral identification tools and offer a blueprint for feature engineering of new tools that will lead to higher-confidence viral discovery in microbiome studies.

Nucleic Transformer: Classifying DNA Sequences with Self-Attention and Convolutions

Much work has been done to apply machine learning and deep learning to genomics tasks, but these applications usually require extensive domain knowledge, and the resulting models provide very limited interpretability. Here, we present the Nucleic Transformer, a conceptually simple but effective and interpretable model architecture that excels in the classification of DNA sequences. The Nucleic Transformer employs self-attention and convolutions on nucleic acid sequences, leveraging two prominent deep learning strategies commonly used in computer vision and natural language analysis. We demonstrate that the Nucleic Transformer can be trained without much domain knowledge to achieve high performance in Escherichia coli promoter classification, viral genome identification, enhancer classification, and chromatin profile predictions.

DETIRE: a hybrid deep learning model for identifying viral sequences from metagenomes

A metagenome contains all DNA sequences from an environmental sample, including viruses, bacteria, archaea, and eukaryotes. Since viruses are of huge abundance and have caused vast mortality and morbidity to human society in history as a type of major pathogens, detecting viruses from metagenomes plays a crucial role in analyzing the viral component of samples and is the very first step for clinical diagnosis. However, detecting viral fragments directly from the metagenomes is still a tough issue because of the existence of a huge number of short sequences. In this study a hybrid Deep lEarning model for idenTifying vIral sequences fRom mEtagenomes (DETIRE) is proposed to solve the problem. First, the graph-based nucleotide sequence embedding strategy is utilized to enrich the expression of DNA sequences by training an embedding matrix. Then, the spatial and sequential features are extracted by trained CNN and BiLSTM networks, respectively, to enrich the features of short sequences. Finally, the two sets of features are weighted combined for the final decision. Trained by 220,000 sequences of 500 bp subsampled from the Virus and Host RefSeq genomes, DETIRE identifies more short viral sequences (<1,000 bp) than the three latest methods, such as DeepVirFinder, PPR-Meta, and CHEER. DETIRE is freely available at Github (https://github.com/crazyinter/DETIRE).

Evaluation of computational phage detection tools for metagenomic datasets

Introduction

As new computational tools for detecting phage in metagenomes are being rapidly developed, a critical need has emerged to develop systematic benchmarks.

Methods

In this study, we surveyed 19 metagenomic phage detection tools, 9 of which could be installed and run at scale. Those 9 tools were assessed on several benchmark challenges. Fragmented reference genomes are used to assess the effects of fragment length, low viral content, phage taxonomy, robustness to eukaryotic contamination, and computational resource usage. Simulated metagenomes are used to assess the effects of sequencing and assembly quality on the tool performances. Finally, real human gut metagenomes and viromes are used to assess the differences and similarities in the phage communities predicted by the tools.

Results

We find that the various tools yield strikingly different results. Generally, tools that use a homology approach (VirSorter, MARVEL, viralVerify, VIBRANT, and VirSorter2) demonstrate low false positive rates and robustness to eukaryotic contamination. Conversely, tools that use a sequence composition approach (VirFinder, DeepVirFinder, Seeker), and MetaPhinder, have higher sensitivity, including to phages with less representation in reference databases. These differences led to widely differing predicted phage communities in human gut metagenomes, with nearly 80% of contigs being marked as phage by at least one tool and a maximum overlap of 38.8% between any two tools. While the results were more consistent among the tools on viromes, the differences in results were still significant, with a maximum overlap of 60.65%. Discussion: Importantly, the benchmark datasets developed in this study are publicly available and reusable to enable the future comparability of new tools developed.

Virsearcher: Identifying Bacteriophages from Metagenomes by Combining Convolutional Neural Network and Gene Information

Metagenome sequencing provides an unprecedented opportunity for the discovery of unknown microbes and viruses. A large number of phages and prokaryotes are mixed together in metagenomes. To study the influence of phages on human bodies and environments, it is of great significance to isolate phages from metagenomes. However, it is difficult to identify novel phages because of the diversity of their sequences and the frequent presence of short contigs in metagenomes. Here, virSearcher is developed to identify phages from metagenomes by combining the convolutional neural network (CNN) and the gene information of input sequences. Firstly, an input sequence is encoded in accordance with the different functions of its coding and the non-coding regions and then is converted into word embedding code through a word embedding layer before a convolutional layer. Meanwhile, the hit ratio of the virus genes is combined with the output of the CNN to further improve the performance of the network. The genes used by virSearcher consist of complete and incomplete genes. Experiments on several metagenomes have showed that, compared with others, virSearcher can significantly improve the performance for the identification of short sequences, while maintaining the performance for long ones. The source code of virSearcher is freely available from http://github.com/DrJackson18/virSearcher.

SourceFinder: a Machine-Learning-Based Tool for Identification of Chromosomal, Plasmid, and Bacteriophage Sequences from Assemblies

High-throughput genome sequencing technologies enable the investigation of complex genetic interactions, including the horizontal gene transfer of plasmids and bacteriophages. However, identifying these elements from assembled reads remains challenging due to genome sequence plasticity and the difficulty in assembling complete sequences. In this study, we developed a classifier, using random forest, to identify whether sequences originated from bacterial chromosomes, plasmids, or bacteriophages. The classifier was trained on a diverse collection of 23,211 chromosomal, plasmid, and bacteriophage sequences from hundreds of bacterial species. In order to adapt the classifier to incomplete sequences, each complete sequence was subsampled into 5,000 nucleotide fragments and further subdivided into k-mers. This three-class classifier succeeded in identifying chromosomes, plasmids, and bacteriophages using k-mer distributions of complete and partial genome sequences, including simulated metagenomic scaffolds with minimum performance of 0.939 area under the receiver operating characteristic curve (AUC). This classifier, implemented as SourceFinder, has been made available as an online web service to help the community with predicting the chromosomal, plasmid, and bacteriophage sources of assembled bacterial sequence data (https://cge.food.dtu.dk/services/SourceFinder/). IMPORTANCE Extra-chromosomal genes encoding antimicrobial resistance, metal resistance, and virulence provide selective advantages for bacterial survival under stress conditions and pose serious threats to human and animal health. These accessory genes can impact the composition of microbiomes by providing selective advantages to their hosts. Accurately identifying extra-chromosomal elements in genome sequence data are critical for understanding gene dissemination trajectories and taking preventative measures. Therefore, in this study, we developed a random forest classifier for identifying the source of bacterial chromosomal, plasmid, and bacteriophage sequences.

Explainable deep neural networks for novel viral genome prediction

Viral infection causes a wide variety of human diseases including cancer and COVID-19. Viruses invade host cells and associate with host molecules, potentially disrupting the normal function of hosts that leads to fatal diseases. Novel viral genome prediction is crucial for understanding the complex viral diseases like AIDS and Ebola. While most existing computational techniques classify viral genomes, the efficiency of the classification depends solely on the structural features extracted. The state-of-the-art DNN models achieved excellent performance by automatic extraction of classification features, but the degree of model explainability is relatively poor. During model training for viral prediction, proposed CNN, CNN-LSTM based methods (EdeepVPP, EdeepVPP-hybrid) automatically extracts features. EdeepVPP also performs model interpretability in order to extract the most important patterns that cause viral genomes through learned filters. It is an interpretable CNN model that extracts vital biologically relevant patterns (features) from feature maps of viral sequences. The EdeepVPP-hybrid predictor outperforms all the existing methods by achieving 0.992 mean AUC-ROC and 0.990 AUC-PR on 19 human metagenomic contig experiment datasets using 10-fold cross-validation. We evaluate the ability of CNN filters to detect patterns across high average activation values. To further asses the robustness of EdeepVPP model, we perform leave-one-experiment-out cross-validation. It can work as a recommendation system to further analyze the raw sequences labeled as ‘unknown’ by alignment-based methods. We show that our interpretable model can extract patterns that are considered to be the most important features for predicting virus sequences through learned filters.