(SHOTGUN) METAGENOMICS. Hélène Touzet, CNRS, CRIStAL

Size: px

Start display at page:

Download "(SHOTGUN) METAGENOMICS. Hélène Touzet, CNRS, CRIStAL"

Christal Jacobs
6 years ago
Views:

1 (SHOTGUN) METAGENOMICS Hélène Touzet, CNRS, CRIStAL

2 Shotgun sequencing for community samples Metagenomics potentially sequences all fragmented DNA in a community includes all microorganisms and viruses gives access to all genes across the entire genomes Metatranscriptomics potentially sequences all fragmented RNA in a community activity of the genes

3 Shotgun sequencing versus amplicon sequencing Who is there? more complete taxonomic information no bias due to PCR amplification (primers) access to the full genomes and genes captures genomes which lack amplicon targets, such as viruses What are they doing? functional information analysis of gene functions, metabolic pathways, etc. to identify the functional potential of the community more expensive new challenges in terms of data processing, storage and analysis

4 Content of this lecture Taxonomic analysis Some general ideas, principles and tools Functional analysis Some general ideas, principles and tools Not presented today : Richness, comparative analysis

5 short reads individual reads assembly filtering 16S reads selection of markers all reads one genome per organism possible / not possible? cf amplicons PhyloSift Metaphlan2 MG-rast MEGAN EBI kraken centrifuge... this afternoon taxonomic classification functional analysis

6 Taxonomic classification Input short reads from a single shotgun metagenomic sequencing experiment (FASTA or FASTQ) Output list of detected microbes and their relative abundances

7 Three main approaches 1. focus on a single phylogenetic marker 2. utilization of multiple markers 3. whole information contained in the reads

8 Approach 1 : One marker choice of the phylogenetic marker ubiquitous in the environment showing some differences between species identification of the reads corresponding to the marker processing of the extracted reads direct classification of the raw reads Qiime, MAPseq,... reconstruction of the full sequence of the marker gene before classification Emirge 2011, MATAM 2017

9 Approach 2 : Multiple markers How to choose the markers? all possible genes for all possible genomes redundant information high running-time due to the size of the database not such a good idea selection of a few universal phylogenetics markers phylosift Selection of clade-specific marker sequences Metaphlan2

10 Phylosift 37 families of elite marker genes congruent phylogenetic histories represent about 1% of an average bacterial genome 16S and 18S ribosomal RNA genes mitochondrial gene families eukaryote-specific gene families viral gene families

11 Metaphlan2 Metagenomic Phylogenetic Analysis successor of Metaphlan (2012, Human Microbiome Project) markers and quasi-markers coding sequences that unequivocally identify specific microbial clades at the species level or higher taxonomic levels markers : specific of the clade quasi-markers : show a minimal number of sequence hits in genomes outside the clade

13 pre-computed database of markers and pseudo markers + clades (LCA in the taxonomy) short read bacteria : 770,000 markers + 130,000 pseudomarkers from 13,000 genomes archaea : 460,000 markers + 4,600 pseudomarkers from 300 genomes eukaryotes : 22,400 markers + 2,550 pseudomarkes from 110 genomes virus : 38,800 markers + 23,000 pseudomarkers from 3500 genomes

14 Metaphlan2 pipeline 1. mapping of short reads on the marker database (Bowtie2) 2. calculation of the relative abundance of each taxonomic unit priority to (strict) markers quasi-markers are added only if the number of (strict) markers is < 200 normalization of the total number of reads in each clade by the nucleotide length of its markers 3. unclassified subclades : reads belonging to clades with no available sequenced genomes are reported as an unclassified subclade of the closest ancestor for which there is available sequence data

15 SampleID Metaphlan2_Analysis k Bacteria k Bacteria p Firmicutes k Bacteria p Bacteroidetes k Bacteria p Actinobacteria k Bacteria p Firmicutes c Negativicutes k Bacteria p Bacteroidetes c Bacteroidia k Bacteria p Firmicutes c Bacilli k Bacteria p Actinobacteria c Actinobacteria k Bacteria p Firmicutes c Negativicutes o Selenomonadales k Bacteria p Bacteroidetes c Bacteroidia o Bacteroidales k Bacteria p Firmicutes c Bacilli o Lactobacillales k Bacteria p Actinobacteria c Actinobacteria o Actinomycetales k Bacteria p Firmicutes c Negativicutes o Selenomonadales f Veillon k Bacteria p Bacteroidetes c Bacteroidia o Bacteroidales f Prevotel k Bacteria p Firmicutes c Bacilli o Lactobacillales f Streptococcac k Bacteria p Actinobacteria c Actinobacteria o Actinomycetales f Ac Kingdom Phylum Class Order Family Genus Species Strain

16 SampleID Metaphlan2_Analysis k Bacteria k Bacteria p Firmicutes k Bacteria p Bacteroidetes k Bacteria p Actinobacteria k Bacteria p Firmicutes c Negativicutes k Bacteria p Bacteroidetes c Bacteroidia k Bacteria p Firmicutes c Bacilli k Bacteria p Actinobacteria c Actinobacteria k Bacteria p Firmicutes c Negativicutes o Selenomonadales k Bacteria p Bacteroidetes c Bacteroidia o Bacteroidales k Bacteria p Firmicutes c Bacilli o Lactobacillales k Bacteria p Actinobacteria c Actinobacteria o Actinomycetales k Bacteria p Firmicutes c Negativicutes o Selenomonadales f Veillon k Bacteria p Bacteroidetes c Bacteroidia o Bacteroidales f Prevotel k Bacteria p Firmicutes c Bacilli o Lactobacillales f Streptococcac k Bacteria p Actinobacteria c Actinobacteria o Actinomycetales f Ac Kingdom Phylum Class Order Family Genus Species Strain

17 Installation and usage local installation only (python) tutorial hclust2 : construction of heatmap graphlan : visualisation with cladograms

18 Approach 3 : Raw information reference genomes + taxonomy no annotation, no phylogenetic markers no alignment between the read and the database GTGCATAGTGAAAACGGAGTTGCTGACGACGAAAGCGACATTGGGATCTGTCTGTTGTCATTCGCGG CGAGGCGGACACTGATTGACACGGTTTTGCAGAAGGTTAGGGGAATAGGTTAAATTGAGTGGCTTAA GGATTAAAGTGTAGTAAACTGTGATTAACGGAGACGGTTTTAAGACAGGAGTTCGCAAAATCAAGCG GTTATTCCTGGTGGTTTAGGCGTACAATGTCCTGAAGAATATTTAAGAAAAAAGCACCCCTCGTCGC CGCGGTCGACCATACCTTCGATTATCGCGCCCACTCTCCCATTAGTCGGCAGAGGTGGTTGTGTTGC ATATTCTAAGGCGTTACGCTGATGAATATTCTACGGAATTGCCATAGGCGTTGAACGCTACACGGAC TATAGAGCGGGTCATCGAAAGGTTATACTCTTGTAGTTAACATGTAGCCCGGCCCTATTAGTACAGC CATTCTCATTATTAAATTTTCTCTACAGCCAAACGACCAAGTGCATTTCCACGGAGCGCGATGGAGA AGCTCTGTAATAGGGACTAAAAGAGTGATGATAATCATGAGTGCCGCGTTATGGTGGTGTCGGAACA CCAGTCGTATTCCTTCTCGAGTTCCGTCCAGTTAAGCGTGACACTCCCAGTGTACCCGCAAACCGTG AGTCAATCGCATGTAGGATGGTCTCCAGACACCGGGGCACCAGTTTTCACGCCCAAAGCATAAACGA AGTCTTAGAACTGGACGTGCCGTTTCTCTGCGAATAATACCTCGAGCTGTACCGTTGTTGCGCTGCC CTCTTATCACATTTGCTTCGACGACTGC

19 Kraken exact k-mer matching complete bacterial, archaeal, and viral genomes in RefSeq NCBI Wood and Salzberg Genome Biology 2014, 15:R46 METHOD Kraken: ultrafast metagenomic sequence classification using exact alignments Derrick E Wood 1,2* and Steven L Salzberg 2,3 Open Access Abstract

20 all 31-mers present in the database + LCA (lowest common ancestor) AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAC AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAG AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAT AAAAAAAAAAAAAAAAAAAAAAAAAAAAACA e9 disctinct k-mers (oct 2017) << 4 31 = 4.6e18 all completed microbial genomes of the RefSeq database 47,768 bacteria + 1,034 archaea + 7,530 viruses Precomputed database

21 1. short read overlapping k-mers 3. assignation of the read 2. identification of the LCA in the taxonomy for each k-mer Read assignation

22 Performances of Kraken very fast excellent results with known/poor results with unknow species high memory demanding 500 GB of disk space to build the database 200 GB to store it Minikraken : reduced databases DB 4GB : 2.7% of k-mers from the original database DB 8GB : 5% of k-mers from the original database Centrifuge : space-efficient evolution of Kraken Burrows-Wheeler Transform

23 Installation and usage kraken : local installation only (C++) centrifuge : local installation only (C++) command line interface

24 Similar tools following the same paradigm LMAT, 2013 Scalable metagenomic taxonomy classification using a reference genome database. Ames SK, Hysom DA, Gardner SN, Lloyd GS, Gokhale MB, Allen JE. Bioinformatics, doi : /bioinformatics/btt389 Clark, 2015 CLARK : fast and accurate classification of metagenomic and genomic sequences using discriminative k-mers R. Ounit, S. Wanamaker, T.J. Close, S. Lonardi BMC Genomics ; 16(1) : 236 One codex (commercial, free demo version) web server based on kraken algorithm registration required Kaiju, 2017 both web server and local installation (CLI) Menzel, P. et al. (2016) Fast and sensitive taxonomic classification for metagenomics with Kaiju. Nat. Commun. 7 :11257

26 Functional classification Input short reads from a single shotgun metagenomic sequencing experiment (FASTA or FASTQ) Output functions of the genes present in the community

27 How to annotate mixtures of short reads? ab initio approaches prediction of coding regions in short reads FragGeneScan homology-based approaches alignment of short reads to a large database of annotated sequences choice of the alignment tool, DNA /protein : BlastX, Diamond,... choice of the database : Eggnog, SEEDS, KEGG, Interpro,... built-in pipelines : MG-RAST, MEGAN, EBI,...

28 FragGeneScan predict protein-coding regions from environmental sequences HMM combining codon usage bias + start/stop codon models (like Glimmer or Genemark) sequencing error models included in the MG-Rast and EBI pipelines

29 Alignment tools Query : large set of short DNA reads Reference : large protein database Pre-NGS tools : BlastX, BLAT Diamond optimized to deal with sort reads order of magnitude faster than BlastX for this kind of data ( 1000)

30 Interpro developed at EBI since 1999 signatures for protein families, domains and functional sites collected from 14 databases mappings of InterPro entries to Gene Ontology (GO) terms (InterPro2GO)

31 KEGG collection of databases : genomes, genes, metabolic pathways, diseases,... KO entries : group of genes representing functional orthologs in the molecular networks

32 1 Structural and Computational Biology Unit, European Molecular Biology Laboratory, Heidelberg, Germany, 2 Institute of Molecular Life Sciences, University of Zurich, Zurich 8057, Switzerland, 3 Bioinformatics/Systems Biology Group, Swiss Institute of Bioinformatics (SIB), Zurich 8057, Switzerland, 4 The Novo Nordisk Foundation Center for Protein Research, Faculty of Health and Medical Sciences, University of Copenhagen, Copenhagen N 2200, Denmark, 5 Institute of Bioinformatics and Systems Biology, Helmholtz Zentrum München, German Research Center for Environmental Health (GmbH), Neuherberg 85764, Germany, 6 CUBE Division of Computational Systems Biology, EggNOG database of orthologous groups of proteins at different taxonomic levels + functional annotation clustering of the 9.6 million proteins from 2031 genomes viruses OG (orthologous groups) D286 D293 Nucleic Acids Research, 2016, Vol. 44, Database issue Published online 17 November 2015 doi: /nar/gkv1248 eggnog 4.5: a hierarchical orthology framework with improved functional annotations for eukaryotic, prokaryotic and viral sequences Jaime Huerta-Cepas 1, Damian Szklarczyk 2,3,KristofferForslund 1,HelenCook 4, Davide Heller 2,3,MathiasC.Walter 5,ThomasRattei 6,DanielR.Mende 7, Shinichi Sunagawa 1,MichaelKuhn 8,LarsJuhlJensen 4,ChristianvonMering 2,3,* and Peer Bork 1,9,10,*

33 SEEDS initiated in 2003 for genome annotation hierarchical way to organize gene families > 1300 protein families

34 Pipelines for functional analysis aligner EggNOG Interpro KEGG SEED MG-RAST BLAT MEGAN CE BlastX Diamond EBI Interproscan

35 MG-RAST Metagenomics Rapid Annotation using Subsystem Technology

36 MG-RAST developed since 2007 (University of Chicago) extension of RAST ( The Project to Annotate 1000 Genomes) Rapid Annotation using Subsystems Technology subsystems : SEED framework also supports amplicons (16S, 18S, and ITS) and metatranscriptomics BMC Bioinformatics BioMed Central Software Open Access The metagenomics RAST server a public resource for the automatic phylogenetic and functional analysis of metagenomes F Meyer* 1,2, D Paarmann 2, M D'Souza 2, R Olson 1, EM Glass 1, M Kubal 2, T Paczian 1, A Rodriguez 2, R Stevens 1,2, A Wilke 2, J Wilkening 1 and RA Edwards 1,3 Address: 1 Mathematics and Computer Science Division, Argonne National Laboratory, 9700 S Cass Avenue, Argonne, IL 60439, USA, 2 3

For shotgun metagenome and shotgun metatranscriptome datasets we perform a dereplication step. Figure 3.1: Details of the analysis pipeline for MG-RAST version 3 cleaning of the sequencing reads 3.

37 For shotgun metagenome and shotgun metatranscriptome datasets we perform a dereplication step. Figure 3.1: Details of the analysis pipeline for MG-RAST version 3 cleaning of the sequencing reads 3.1 Data hygiene rrna detection : SortmeRNA/ Silva + RDP classifier Preprocessing Protein coding gene calling : FragGeneScan (prokaryotes) After upload, data preprocessed by using SolexaQA [8] to trim low-quality regions from FASTQ data. Comparison Platform-specific toapproaches GenBank, are usedseed, for 454 data IMG, submitted UniProt, in FASTA format: KEGG, readsand more than eggnogs than two standard withdeviations BLATaway from the mean read length are discarded following [14]. All sequences submitted to the system are available, but discarded reads will not be analyzed further. Dereplication

38 Usage easy accessibility via a web-based interface account identification storage on the MG-rast server sequencing data + metadata status : private, public,... priority queue depending on the status, may be very slow 315,470 metagenomes containing 1,147 billion sequences and Tbp processed for 24,415 registered users.

41 MEGAN

42 MEGAN developed since 2007 (U. Tübingen) last release : MEGAN CE, 2017

43 MEGAN Community Edition - Interactive Exploration and Analysis of Microbiome Sequencing Data Fig 1. The compressed fastq files in a metagenome or metatranscriptome sequencing project are (1) compared against a protein reference database such as NCBI-nr using DIAMOND. (2) Taxonomic and functional analysis is then performed on the diamond files using Meganizer. (3) The resulting meganized diamond files remain on the server and are accessed via the MeganServer software. (4) Researchers work interactively with the data using MEGAN CE. Initial database : NCBI nr + NCBI taxonomy doi: /journal.pcbi g001 Design and Implementation Alignment of the reads on the database : Diamond This paper introduces MEGAN Community Edition (CE), which is a major update of our MEGAN software [11]. This release contains a large number of new features and has been substantially rewritten so as to support the analysis of many samples (hundreds) and many reads Exploration of the (billions). data This release (MEGAN includes a number of CE) command: line tools, in particular blast2rma, daa2rma and Meganizer, which can all be used to prepare input files for MEGAN CE. Taxonomic binning In addition, we : recommend LCA, the lowest use of DIAMOND common [10] for ultra-fast ancestor alignment of reads against NCBI-nr and MeganServer to allow web access to MEGAN files. Functional analysis : mapping to KEGG, SEED, EggNOG and RMA files InterPro2GO MEGAN CE analysisrequires that sequencing reads are first aligned against a suitable reference database, such as NCBI-nr in the case of a protein-based analysis, or Genbank for a DNA-based analysis, or the Silva database [24], say, when aligning 16S rrna reads. MEGAN CE can import readsand alignments in a number of different file formats and computes a compressed and indexed binary file in so-called RMA format that contains all reads, alignments, taxonomic and functional classifications. The file isindexed to allow quick access to reads and alignments bytax-

44 Lowest common ancestor with MEGAN (2007) origin of the LCA

45 Usage diamond : local installation MEGAN CE (Community Edition) : local installation JAVA graphical interface multiple viewers : KEGG Viewer (pathways), EggNog viewer, comparative analysis, rarefaction,... MEGAN Ultimate Edition

46 EBI Metagenomics

47 EBI Metagenomics first public release in 2013 close integration with the ENA (European Nucleotide Archive)

48 EBI Metagenomics pipeline cleaning and trimming of the short reads taxonomic analysis : Qiime, and then Mapseq on 16S reads functional analysis : InterPro + InterProScan + InterPro2GO

49 Data submission : ENA Webin data are stably archived accession numbers (prerequisite for many publications) active submission helpdesk training materials A major aim in the development of this resource has been to encourage metagenomics researchers to openly share their data as widely as possible, and to also describe their data in sufficient detail such that other scientists are able to extract maximum value from it.

50 8 Nucleic Acids Research, 2017 Figure 6. Growth of metagenomics data housed in ENA and processed by EBI Metagenomics (EMG). This graph shows the cumulative growth of environmental data in the two resources (ENA: solid lines, EMG: dashed lines) according to two different metrics: numbers of samples (blue) and number of bases (orange). to export the documents in numerous formats and to support concurrent versions for different pipeline/website releases. Furthermore, as this documentation is available in GitHub, it allows team members, external collaborators and users to contribute to the documentation serving contextual metadata and results have been developed to offer powerful entry points for browsing, searching and discovery of data from both a manual and programmatic perspective. Amorefundamentalchangeistheshifttowardsprovision of assembly of both user-submitted and publicly avail-

51 Usage webserver : upload of the data, analyses on the cloud programmatic access : REST API krona visualisation maybe very slow (several days)

52 Conclusion fast evolving field influence of the nature of the data sequencing technology and quality of the data complexity of the community coverage balance between performances and usability

(including CLARK, Kraken, LMAT, Metaphlan2, PhyloSift,

53 Which tool is the best? With which parameters? 2016, MEGAN (older version), MG-RAST, One Codex 2017, 11 tools (including CLARK, Kraken, LMAT, Metaphlan2, PhyloSift, MGAN+Diamond) 2016, 14 tools (including CLARK, MetaPhan2, One codex, EBI, MG-Rast, kraken, LMAT, Megan)

54 Shotgun sequencing versus amplicon sequencing Comparing 16S rrna Marker Gene and Shotgun Metagenomics Datasets in the American Gut Project Using State of the Art Tools, E.R. Hyde, J. Sanders, A. Tripathi, Q. Zhu, R. Knight, 2017 There is some consistency between the 16S and shotgun metagenomics approaches although some obvious differences are noted. Large-scale differences in microbial biodiversity discovery between 16S amplicon and shotgun sequencing, Michael Tessler et al., Scientific Reports 2017 Overall the amplicon data were more robust across both biodiversity and community ecology analyses at different taxonomic scales.

Introduction to taxonomic analysis of metagenomic amplicon and shotgun data with QIIME. Peter Sterk EBI Metagenomics Course 2014

Introduction to taxonomic analysis of metagenomic amplicon and shotgun data with QIIME Peter Sterk EBI Metagenomics Course 2014 1 Taxonomic analysis using next-generation sequencing Objective we want to