Title: Learning, visualizing and exploring 16S rRNA structure using an attention-based deep neural network
Recurrent neural networks with memory and attention mechanisms are widely used in natural language processing because they can capture short and long term sequential information for diverse tasks. We propose an integrated deep learning model for microbial DNA sequence data, which exploits convolutional neural networks, recurrent neural networks, and attention mechanisms to predict taxonomic classifications and sample-associated attributes, such as the relationship between the microbiome and host phenotype, on the read/sequence level. In this paper, we develop this novel deep learning approach and evaluate its application to amplicon sequences. We apply our approach to short DNA reads and full sequences of 16S ribosomal RNA (rRNA) marker genes, which identify the heterogeneity of a microbial community sample. We demonstrate that our implementation of a novel attention-based deep network architecture, Read2Pheno , achieves read-level phenotypic prediction. Training Read2Pheno models will encode sequences (reads) into dense, meaningful representations: learned embedded vectors output from the intermediate layer of the network model, which can provide biological insight when visualized. The attention layer of Read2Pheno models can also automatically identify nucleotide regions in reads/sequences which are particularly informative for classification. As such, this novel approach can avoid pre/post-processing and manual interpretation required with conventional approaches to microbiome sequence classification. We further show, as proof-of-concept, that aggregating read-level information can robustly predict microbial community properties, host phenotype, and taxonomic classification, with performance at least comparable to conventional approaches. An implementation of the attention-based deep learning network is available at https://github.com/EESI/sequence_attention (a python package) and https://github.com/EESI/seq2att (a command line tool). more »« less
Saha, Subrata; Johnson, Jethro; Pal, Soumitra; Weinstock, George M.; Rajasekaran, Sanguthevar; Hancock, ed., John(
, Bioinformatics)
AbstractMotivation
Metagenomics is the study of genetic materials directly sampled from natural habitats. It has the potential to reveal previously hidden diversity of microscopic life largely due to the existence of highly parallel and low-cost next-generation sequencing technology. Conventional approaches align metagenomic reads onto known reference genomes to identify microbes in the sample. Since such a collection of reference genomes is very large, the approach often needs high-end computing machines with large memory which is not often available to researchers. Alternative approaches follow an alignment-free methodology where the presence of a microbe is predicted using the information about the unique k-mers present in the microbial genomes. However, such approaches suffer from high false positives due to trading off the value of k with the computational resources. In this article, we propose a highly efficient metagenomic sequence classification (MSC) algorithm that is a hybrid of both approaches. Instead of aligning reads to the full genomes, MSC aligns reads onto a set of carefully chosen, shorter and highly discriminating model sequences built from the unique k-mers of each of the reference sequences.
Results
Microbiome researchers are generally interested in two objectives of a taxonomic classifier: (i) to detect prevalence, i.e. the taxa present in a sample, and (ii) to estimate their relative abundances. MSC is primarily designed to detect prevalence and experimental results show that MSC is indeed a more effective and efficient algorithm compared to the other state-of-the-art algorithms in terms of accuracy, memory and runtime. Moreover, MSC outputs an approximate estimate of the abundances.
Availability and implementation
The implementations are freely available for non-commercial purposes. They can be downloaded from https://drive.google.com/open?id=1XirkAamkQ3ltWvI1W1igYQFusp9DHtVl.
Advances in microbiome science are being driven in large part due to our ability to study and infer microbial ecology from genomes reconstructed from mixed microbial communities using metagenomics and single-cell genomics. Such omics-based techniques allow us to read genomic blueprints of microorganisms, decipher their functional capacities and activities, and reconstruct their roles in biogeochemical processes. Currently available tools for analyses of genomic data can annotate and depict metabolic functions to some extent; however, no standardized approaches are currently available for the comprehensive characterization of metabolic predictions, metabolite exchanges, microbial interactions, and microbial contributions to biogeochemical cycling.
Results
We present METABOLIC (METabolic And BiogeOchemistry anaLyses In miCrobes), a scalable software to advance microbial ecology and biogeochemistry studies using genomes at the resolution of individual organisms and/or microbial communities. The genome-scale workflow includes annotation of microbial genomes, motif validation of biochemically validated conserved protein residues, metabolic pathway analyses, and calculation of contributions to individual biogeochemical transformations and cycles. The community-scale workflow supplements genome-scale analyses with determination of genome abundance in the microbiome, potential microbial metabolic handoffs and metabolite exchange, reconstruction of functional networks, and determination of microbial contributions to biogeochemical cycles. METABOLIC can take input genomes from isolates, metagenome-assembled genomes, or single-cell genomes. Results are presented in the form of tables for metabolism and a variety of visualizations including biogeochemical cycling potential, representation of sequential metabolic transformations, community-scale microbial functional networks using a newly defined metric “MW-score” (metabolic weight score), and metabolic Sankey diagrams. METABOLIC takes ~ 3 h with 40 CPU threads to process ~ 100 genomes and corresponding metagenomic reads within which the most compute-demanding part of hmmsearch takes ~ 45 min, while it takes ~ 5 h to complete hmmsearch for ~ 3600 genomes. Tests of accuracy, robustness, and consistency suggest METABOLIC provides better performance compared to other software and online servers. To highlight the utility and versatility of METABOLIC, we demonstrate its capabilities on diverse metagenomic datasets from the marine subsurface, terrestrial subsurface, meadow soil, deep sea, freshwater lakes, wastewater, and the human gut.
Conclusion
METABOLIC enables the consistent and reproducible study of microbial community ecology and biogeochemistry using a foundation of genome-informed microbial metabolism, and will advance the integration of uncultivated organisms into metabolic and biogeochemical models. METABOLIC is written in Perl and R and is freely available under GPLv3 athttps://github.com/AnantharamanLab/METABOLIC.
Dietary DNA metabarcoding enables researchers to identify and characterize trophic interactions with a high degree of taxonomic precision. It is also sensitive to sources of bias and contamination in the field and lab. One of the earliest and most common strategies for dealing with such sensitivities has been to filter resulting sequence data to remove low-abundance sequences before conducting ecological analyses based on the presence or absence of food taxa. Although this step is now often perceived to be both necessary and sufficient for cleaning up datasets, evidence to support this perception is lacking and more attention needs to be paid to the related risk of introducing other undesirable errors. Using computer simulations, we demonstrate that common strategies to remove low-abundance sequences can erroneously eliminate true dietary sequences in ways that impact downstream dietary inferences. Using real data from well-studied wildlife populations in Yellowstone National Park, we further show how these strategies can markedly alter the composition of individual dietary profiles in ways that scale-up to obscure ecological interpretations about dietary generalism, specialism, and niche partitioning. Although the practice of removing low-abundance sequences may continue to be a useful strategy to address a subset of research questions that focus on a subset of relatively abundant food resources, its continued widespread use risks generating misleading perceptions about the structure of trophic networks. Researchers working with dietary DNA metabarcoding data—or similar data such as environmental DNA, microbiomes, or pathobiomes—should be aware of potential drawbacks and consider alternative bioinformatic, experimental, and statistical solutions. We used fecal DNA metabarcoding to characterize the diets of bison and bighorn sheep in winter and summer. Our analyses are based on 35 samples (median per species per season = 10) analyzed using the P6 loop of the chloroplast trnL(UAA) intron together with publicly available plant reference data (Illumina sequence read data are available at NCBI (BioProject: PRJNA780500)). Obicut was used to trim reads with a minimum quality threshold of 30, and primers were removed from forward and reverse reads using cutadapt. All further sequence identifications were performed using obitools; forward and reverse sequences were aligned using the illuminapairedend command using a minimum alignment score of 40, and only joined sequences retained. We used the obiuniq command to group identical sequences and tally them within samples, enabling us to quantify the relative read abundance (RRA) of each sequence. Sequences that occurred ≤2 times overall or that were ≤8 bp were discarded. Sequences were considered to be likely PCR artifacts if they were highly similar to another sequence (1 bp difference) and had a much lower abundance (0.05%) in the majority of samples in which they occurred; we discarded these sequences using the obiclean command. Overall, we characterized 357 plant sequences and a subset of 355 sequences were retained in the dataset after rarefying samples to equal sequencing depth. We then applied relative read abundance thresholds from 0% to 5% to the fecal samples. We compared differences in the inferred dietary richness within and between species based on individual samples, based on average richness across samples, and based on the total richness of each population after accounting for differences in sample size. The readme file contains an explanation of each of the variables in the dataset. Information on the methodology can be found in the associated manuscript referenced above.
Abstract Motivation Accurate prediction of residue-residue distances is important for protein structure prediction. We developed several protein distance predictors based on a deep learning distance prediction method and blindly tested them in the 14th Critical Assessment of Protein Structure Prediction (CASP14). The prediction method uses deep residual neural networks with the channel-wise attention mechanism to classify the distance between every two residues into multiple distance intervals. The input features for the deep learning method include co-evolutionary features as well as other sequence-based features derived from multiple sequence alignments (MSAs). Three alignment methods are used with multiple protein sequence/profile databases to generate MSAs for input feature generation. Based on different configurations and training strategies of the deep learning method, five MULTICOM distance predictors were created to participate in the CASP14 experiment. Results Benchmarked on 37 hard CASP14 domains, the best performing MULTICOM predictor is ranked 5th out of 30 automated CASP14 distance prediction servers in terms of precision of top L/5 long-range contact predictions (i.e. classifying distances between two residues into two categories: in contact (< 8 Angstrom) and not in contact otherwise) and performs better than the best CASP13 distance prediction method. The best performing MULTICOM predictor is also ranked 6th among automated server predictors in classifying inter-residue distances into 10 distance intervals defined by CASP14 according to the precision of distance classification. The results show that the quality and depth of MSAs depend on alignment methods and sequence databases and have a significant impact on the accuracy of distance prediction. Using larger training datasets and multiple complementary features improves prediction accuracy. However, the number of effective sequences in MSAs is only a weak indicator of the quality of MSAs and the accuracy of predicted distance maps. In contrast, there is a strong correlation between the accuracy of contact/distance predictions and the average probability of the predicted contacts, which can therefore be more effectively used to estimate the confidence of distance predictions and select predicted distance maps. Availability The software package, source code, and data of DeepDist2 are freely available at https://github.com/multicom-toolbox/deepdist and https://zenodo.org/record/4712084#.YIIM13VKhQM.
Indexing reference sequences for search—both individual genomes and collections of genomes—is an important building block for many sequence analysis tasks. Much work has been dedicated to developing full-text indices for genomic sequences, based on data structures such as the suffix array, the BWT and the FM-index. However, the de Bruijn graph, commonly used for sequence assembly, has recently been gaining attention as an indexing data structure, due to its natural ability to represent multiple references using a graphical structure, and to collapse highly-repetitive sequence regions. Yet, much less attention has been given as to how to best index such a structure, such that queries can be performed efficiently and memory usage remains practical as the size and number of reference sequences being indexed grows large.
Results
We present a novel data structure for representing and indexing the compacted colored de Bruijn graph, which allows for efficient pattern matching and retrieval of the reference information associated with each k-mer. As the popularity of the de Bruijn graph as an index has increased over the past few years, so have the number of proposed representations of this structure. Existing structures typically fall into two categories; those that are hashing-based and provide very fast access to the underlying k-mer information, and those that are space-frugal and provide asymptotically efficient but practically slower pattern search. Our representation achieves a compromise between these two extremes. By building upon minimum perfect hashing and making use of succinct representations where applicable, our data structure provides practically fast lookup while greatly reducing the space compared to traditional hashing-based implementations. Further, we describe a sampling scheme for this index, which provides the ability to trade off query speed for a reduction in the index size. We believe this representation strikes a desirable balance between speed and space usage, and allows for fast search on large reference sequences.
Finally, we describe an application of this index to the taxonomic read assignment problem. We show that by adopting, essentially, the approach of Kraken, but replacing k-mer presence with coverage by chains of consistent unique maximal matches, we can improve the space, speed and accuracy of taxonomic read assignment.
Availability and implementation
pufferfish is written in C++11, is open source, and is available at https://github.com/COMBINE-lab/pufferfish.
Supplementary information
Supplementary data are available at Bioinformatics online.
Zhao, Zhengqiao, Woloszynek, Stephen, Agbavor, Felix, Mell, Joshua Chang, Sokhansanj, Bahrad A., and Rosen, Gail L. Learning, visualizing and exploring 16S rRNA structure using an attention-based deep neural network. Retrieved from https://par.nsf.gov/biblio/10356868. PLOS Computational Biology 17.9 Web. doi:10.1371/journal.pcbi.1009345.
Zhao, Zhengqiao, Woloszynek, Stephen, Agbavor, Felix, Mell, Joshua Chang, Sokhansanj, Bahrad A., and Rosen, Gail L.
"Learning, visualizing and exploring 16S rRNA structure using an attention-based deep neural network". PLOS Computational Biology 17 (9). Country unknown/Code not available. https://doi.org/10.1371/journal.pcbi.1009345.https://par.nsf.gov/biblio/10356868.
@article{osti_10356868,
place = {Country unknown/Code not available},
title = {Learning, visualizing and exploring 16S rRNA structure using an attention-based deep neural network},
url = {https://par.nsf.gov/biblio/10356868},
DOI = {10.1371/journal.pcbi.1009345},
abstractNote = {Recurrent neural networks with memory and attention mechanisms are widely used in natural language processing because they can capture short and long term sequential information for diverse tasks. We propose an integrated deep learning model for microbial DNA sequence data, which exploits convolutional neural networks, recurrent neural networks, and attention mechanisms to predict taxonomic classifications and sample-associated attributes, such as the relationship between the microbiome and host phenotype, on the read/sequence level. In this paper, we develop this novel deep learning approach and evaluate its application to amplicon sequences. We apply our approach to short DNA reads and full sequences of 16S ribosomal RNA (rRNA) marker genes, which identify the heterogeneity of a microbial community sample. We demonstrate that our implementation of a novel attention-based deep network architecture, Read2Pheno , achieves read-level phenotypic prediction. Training Read2Pheno models will encode sequences (reads) into dense, meaningful representations: learned embedded vectors output from the intermediate layer of the network model, which can provide biological insight when visualized. The attention layer of Read2Pheno models can also automatically identify nucleotide regions in reads/sequences which are particularly informative for classification. As such, this novel approach can avoid pre/post-processing and manual interpretation required with conventional approaches to microbiome sequence classification. We further show, as proof-of-concept, that aggregating read-level information can robustly predict microbial community properties, host phenotype, and taxonomic classification, with performance at least comparable to conventional approaches. An implementation of the attention-based deep learning network is available at https://github.com/EESI/sequence_attention (a python package) and https://github.com/EESI/seq2att (a command line tool).},
journal = {PLOS Computational Biology},
volume = {17},
number = {9},
author = {Zhao, Zhengqiao and Woloszynek, Stephen and Agbavor, Felix and Mell, Joshua Chang and Sokhansanj, Bahrad A. and Rosen, Gail L.},
editor = {Borenstein, Elhanan}
}
Warning: Leaving National Science Foundation Website
You are now leaving the National Science Foundation website to go to a non-government website.
Website:
NSF takes no responsibility for and exercises no control over the views expressed or the accuracy of
the information contained on this site. Also be aware that NSF's privacy policy does not apply to this site.
