http www ebi ac ukmetagenomics Hubert DENISE hudeniseebi

http: //www. ebi. ac. uk/metagenomics Hubert DENISE hudenise@ebi. ac. uk

1997 Ph. D. Molecular Parasitology Univ. Bordeaux II, France About me 2003 – 2005 Lecturer Molecular Biology, Univ. Clermont-Ferrand II, France 1997 - 2003 Post. Doc, WCMP Univ. Glasgow, UK 2011 – 2012 MSc. Bioinformatics Univ. Cranfield, UK 2005 - 2011 Sr. Scientist, Pfizer Ltd Sandwich, UK 2012 Bioinformatician Sanger Institute then EBI, Hinxton, UK

Where is the true cost of NGS ? 14. 5 % 30 % 28 % 70 % (~80 bp/$) (~2 m bp/$) 4. 5 % 14. 5 % 55 % 36. 5 % 14. 5 % Sboner et al. Genome Biology (2011) 12: 125

EBI Metagenomics pipeline Data analysis using selected EBI and external software tools § Philosophy § Submission to EBI Metagenomics § § QC steps Overview of functional analysis Overview of taxonomy analysis Metagenome assembly § Result outputs § Others public pipelines

Philosophy behind EBI Metagenomics pipeline Helping metagenomics researchers make sense of their data From chaos to structure: § archiving of data with metadata § performing stringent QC filtering prior to analysis § quality in, quality out § performing robust taxonomy and functional analysis § model-based rather than similarity-based approaches § assignment done on reads rather than assembly § intuitive navigation through website § constant drive to improvement § benchmarking and tool testing

EBI Metagenomics pipeline Data analysis using selected EBI and external software tools § Philosophy § Submission to EBI Metagenomics § § QC steps Overview of functional analysis Overview of taxonomy analysis Metagenome assembly § Result outputs § Others public pipelines

http: //www. ebi. ac. uk/metagenomics secure login Navigation panes Resource stats Latest data and news

Submitting to EBI Metagenomics • Your data is valuable to you • Raw sequence data • Description of sample and experiment (sample metadata) • Analysis steps and results • All of this needs to be captured and stored to give context to your data • If so, your data can also be valuable to others

Submitting to EBI Metagenomics • EBI Metagenomics want to encourage people to supply as much detailed metadata as possible, but with the lowest possible overhead who where, when, what how • Development of intuitive web-based tools : ENA Webin and ISA tools • Use of templates and check-lists (MIGS/MIXS standards) • Tutorial and direct support

EBI Metagenomics pipeline Data analysis using selected EBI and external software tools § Phylosophy § Submission to EBI Metagenomics § § QC steps Overview of functional analysis Overview of taxonomy analysis Metagenome assembly § Result outputs § Others public pipelines

Metagenomics data analysis Diversity analysis Quality control Functional analysis Image credits: (1) Christina Toft & Siv G. E. Andersson; (2) Dalebroux Z D et al. Microbiol. Mol. Biol. Rev. 2010; 74: 171 -199

Overview of EBI Metagenomics Pipeline raw reads trim and QC remove short remove duplicates processed reads r. RNAselector discarded reads Amplicon-based data reads with r. RNA reads without r. RNA Frag. Gene. Scan predicted CDS Qiime Inter. Pro. Scan Taxonomic analysis Function assignment Unknown function p. CDS

EBI Metagenomics: QC rationale Why ? § Garbage in, garbage out § Base call error: - each base call has a quality score associated - specific platform-dependent errors § Reads quality decreases with reads length § NGS generates duplicate reads (false and real). Reducing duplication reduces analysis time and prevent analysis bias.

EBI Metagenomics: QC step by step § Clipping - low quality ends trimmed and adapter sequences removed using Biopython Seq. IO package § Quality filtering - sequences with > 10% undetermined nucleotides removed § Read length filtering - short sequences are removed § Duplicate sequences removal - clustered on 99% identity (UCLUST v 1. 1. 579) and representative sequence chosen § Repeat masking - Repeat. Masker (open-3. 2. 2), removed reads with 50% or more nucleotides masked

EBI Metagenomics: QC consequences Roche 454 Ion Torrent Illumina

EBI Metagenomics: overview of functional analysis reads without r. RNA Frag. Gene. Scan predicted CDS Inter. Pro. Scan Function assignment Unknown function p. CDS

EBI Metagenomics: identification of coding sequences Prediction of coding sequences is a challenge § read length § sequencing errors: frame-shift Two main types of approaches: § homology-based methods: identify only known coding sequences § feature-based approaches: predict probability that ORFs are coding EBI Metagenomics uses Frag. Gene. Scan : § hidden Markov models to correct frame-shift using codon usage § probabilistic identification of start and stop codons § 60 bp minimum ORF Rho et al. (2010) NAR 38 -20

EBI Metagenomics: annotation of coding sequences Most available pipelines use pairwise alignment methods (such as BLAST) § compare a query sequence with a database of sequences § identify database sequences that resemble the query sequence with homology score above a certain threshold However sequences may appear to have low homology score because: § proteins may share homology only in limited domains § proteins from different species can differ in length Example: first line of blast alignment of 60 S acidic ribosomal protein P 0 from 2 closely-related species

Using BLAST for annotation

EBI Metagenomics: advantage of Inter. Pro EBI Metagenomics pipeline do not use BLAST-based methods to associate functions to predicted protein sequences: instead we use Inter. Pro. Scan to mine the Inter. Pro database (HMM and profile –based functional analysis) is based on presence of “signatures” (models) from eleven databases § Specificity: mapping is manually curated IPR 024185: 5 -formyltetrahydrofolate cyclo-ligase-like IPR 000847: Transcription regulator HTH, Lys. R § Speed Test set of 40, 692 predicted protein sequences § BLAST vs Uni. Ref 100 = 21. 5 s/cds § Inter. Pro. Scan (5 databases) = 3 s/cds

EBI Metagenomics: Inter. Pro. Scan annotations member database p. CDS signature accession signature description SRR 413626. 9733695_1_1_105_- Pro. Site. Patterns PS 00194 Thioredoxin family active site 1. 0 E-13 IPR 017937 score Inter. Pro accession Thioredoxin, conserved site GO: 0045454 Inter. Pro description GO annotation

EBI Metagenomics: Inter. Pro. Scan annotations signatures links description GO terms

Aims of the Gene Ontology • Controlled vocabulary • Unify the representation of gene and gene product attributes across species • Allow cross-species and/or cross-database comparisons

Inconsistency in naming of biological concepts English is not a very precise language • Same name for different concepts • Different names for the same concept An example … Taction Tactile sense ? Sensory perception of touch ; GO: 0050975

The Gene Ontology Less specific concepts • A way to capture biological knowledge in a written and computable form • A set of concepts and their relationships to each other arranged as a hierarchy More specific concepts www. ebi. ac. uk/Quick. GO

The Concepts in GO 1. Molecular Function • • An elemental activity or task or job protein kinase activity insulin receptor activity 2. Biological Process A commonly recognised series of events • cell division • mitochondrion 3. Cellular Component Where a gene product is located • mitochondrial matrix • mitochondrial inner membrane

The relationship between Inter. Pro and GO (Inter. Pro 2 GO) • Curators manually add relevant GO terms to Inter. Pro entries • When a sequence is searched against Inter. Pro, it is assigned GO terms by virtue of the entries it matches SRR 413626. 11302948_1_1_133_+ 8. 9 E-6 IPR 003439 Pfam PF 00005 ABC transporter 6 ABC transporter-like GO: 0005524|GO: 0016887 ATP binding ATPase activity

EBI Metagenomics: overview of taxonomy analysis processed reads r. RNAselector reads with r. RNA Amplicon-based data Qiime Taxonomic analysis

EBI Metagenomics: identification of suitable sequences Taxonomy analysis is generally based on identification and classification of r. RNA sequences § Prokaryotes: archaebacteria and eubacteria: 5 S, 16 S and 23 S § Eukaryotes: 5 S, 5. 8 S, 18 S and 28 S § there is no equivalent for virus so depend on DNA polymerase or part of 5’-UTR (internal ribosomal entry site [IRES]) sequences EBI Metagenomics currently only provide taxonomy analysis for Prokaryotes. r. RNA sequences are identified using r. RNASelector : § hidden Markov models to identified r. RNA sequences § 60 bp minimum overlap with well-curated HMM model § E-value < 10 -5 Lee et al (2011) J Microbiol. 49(4)

EBI Metagenomics: identification of suitable sequences Once identified, r. RNA sequences are clustered and classified using Qiime “QIIME stands for Quantitative Insights Into Microbial Ecology. QIIME is an open source software package for comparison and analysis of microbial communities” The main steps are: § clustering sequences in Operational Taxonomy Unit (OTU) using uclust § picking a representative sequence set (one sequence from each OTU) § aligning the representative sequence set § assigning taxonomy to the representative sequence set using Py. NAST § generating output files: § filtering the alignment prior to tree building § building phylogenetic tree § creating OTU table

EBI Metagenomics: validation of taxonomy analysis Re-analysis of: Sutton et al, Appl. Environ. Microbiol (2013), 79(2): 619 Impact of Long-Term Diesel Contamination on Soil Microbial Community Structure. Alpha diversity analysis clean polluted clean (outlier)

Assembly of metagenomics data • Metagenomics: Not clear how you avoid assembling sequences from different species together : chimaera • No reference sequence to align against

EBI Metagenomics currently do not perform assembly We are still able to annotate metagenome as show by this re-analysis of Rumen metagenomics by Hess et al, Science (1011) 331: 463 What are the consequences ? § cannot link taxonomy information to functional annotations § cannot currently perform viral taxonomy analysis

EBI Metagenomics pipeline in a nut shell § QC : - trim adaptor sequences, low quality sequence ends - remove duplicates and short sequences - remove low complexity sequences, “Powerful and sophisticated alternative to BLAST-based functional metagenomic analysis” § Diversity analysis : - identify prokaryotic r. RNAsequences (5, 16 and 23 s) - cluster r. RNA-containing reads - assign taxonomy classificationusing Qiime, § Functional analysis : - predict ORFs - translate ORFs into peptides - submit to Inter. Pro. Scan for functional annotation

EBI Metagenomics pipeline Data analysis using selected EBI and external software tools § Submission § § Philosophy Overview data analysis § § QC steps Overview of functional analysis Overview of taxonomy analysis Metagenome assembly § Result outputs § Others public pipelines

Current outputs of EBI Metagenomics pipeline Visualisation Download - QC and sequence statistics - Diversity analysis - Functional analysis

Current outputs of EBI Metagenomics pipeline navigation tabs Access via the Sample page

EBI Metagenomics pipeline: taxonomy visualisation switch to bar chart, column or Krona interactive views Krona interactive representation Google charts dynamic representation

EBI Metagenomics pipeline: functional visualisation Google charts dynamic representation links to Inter. Pro website switch to bar chart view

EBI Metagenomics pipeline : download options 470 MB: need high computing power to manipulate: EBI Metagenomics take care of it and extract meaningful information sets relatively small files: can be manipulated on labtop/desktop computer: users can filtered them according to their needs

EBI Metagenomics pipeline Data analysis using selected EBI and external software tools § Submission § § Philosophy Overview data analysis § § QC steps Overview of functional analysis Overview of taxonomy analysis Metagenome assembly § Result outputs § Others public pipelines

Metagenomics data analysis Quality control Pipeline 1 Taxonomy analysis Functional analysis results 1 results 2 § should share trends and main findings § could differ in ratio and assignment Pipeline 2

Public Metagenomics portals http: //www. ebi. ac. uk/metagenomics/ http: //camera. calit 2. net/ http: //metagenomics. anl. gov/ http: //img. jgi. doe. gov/

Simplified overview of MG-RAST pipeline Sequencer output Abundance profiles • Community reconstruction • Metabolic model http: //metagenomics. anl. gov/ Quality control Similarities search Blat Feature prediction (Frag. Gene. Scan) Clustering (Uclust)

MG-RAST and EBI Metagenomics QC comparison Example: Analysis of Prairie Soil Sample Upload: bp Count Upload: Sequences Count Upload: Mean Sequence Length Upload: Mean GC percent MG-RAST EBI Metagenomics 391, 415, 961 bp 946, 839 413 ± 125 bp 413. 39 bp 61 ± 8 % 61. 2 % Artificial Duplicate Reads: Sequence Count Post QC: bp Count Post QC: Sequences Count Post QC: Mean Sequence Length Post QC: Mean GC percent 0 391, 415, 961 bp 946, 839 413 ± 125 bp 61 ± 8 % 0 388, 670, 692 bp 908, 602 380. 43 bp 57. 8 % 972, 409 5 510, 221 999, 433 3 480, 560 1, 069 1, 110 442, 070 462, 475 Processed: Predicted Protein Features Processed: Predicted r. RNA Features Alignment: Identified Protein Features Alignment: Identified r. RNA Features Annotation: Identified Functional Categories

MG-RAST and EBI Metagenomics Functional analysis Example: Analysis of Prairie Soil Sample ammonia monooxygenase: NH 3 + A-H 2 + O 2 NH 2 OH + A + H 2 O MG-RAST: 28 unique hits on 8 different protein databases 12 2 4 5 62 3 4 1 ammonia monooxygenase family protein Ammonia monooxygenase 2 ammonia monooxygenase subunit A ammonia monooxygenase family protein 1 ammonia monooxygenase, putative Ammonia monooxygenase subunit A 6 putative ammonia monooxygenase Ammonia monooxygenase, putative 2 Putative ammonia monooxygenase 1 putative ammonia monooxygenase subunit A putative ammonia monooxygenase protein putative ammonia monooxygenase subunit A 1 putative ammonia monooxygenase 3 Putative ammonia monooxygenase 5 Ammonia monooxygenase 8 18 13 11 8 10 12 9 9 KEGG egg. NOG Gen. Bank IMG PATRIC Ref. Seq Tr. EMBL SEED what do the abundance numbers mean ? EBI Metagenomics: 3 25 IPR 003393 IPR 007820 Ammonia monooxygenase/particulate methane monooxygenase, subunit A Putative ammonia monooxygenase/protein Abr. B

MG-RAST and EBI Metagenomics Taxonomy analysis Example: Analysis of Prairie Soil Sample MG-RAST domain level of taxonomy Bacteria (55 categories) Archaebacteria (15 categories) Eukaryotes (98 categories) Others (including virus) (3 types) EBI Metagenomics only Archae/Bacteria taxonomy (333 OTU)

Overview of CAMERA workflow

Integrated Microbial Genomes and Metagenomes analysis tools

Some other Metagenomics tools http: //ab. inf. uni-tuebingen. de/software/megan/ http: //www. computationalbioenergy. org/software. html http: //cbcb. umd. edu/software/met. AMOS

Overview of MEGAN rdp, biome files csv, tsv files Comparative visualisation abundance plots PCA, clustering, co-occurrence seq comparison and assignment blast output SAM files csv, tsv files QC ? Taxonomy analysis Functional analysis SEED KEGG COG/EGGNOG

Example of taxonomy analysis using MEGAN diverse single and multi-sample visualisations

Example of taxonomy analysis using MEGAN Comparison, PCA and co-occurrence plots

EBI Metagenomics pipeline Data analysis using selected EBI and external software tools § Submission § § Philosophy Overview data analysis § § QC steps Overview of functional analysis Overview of taxonomy analysis Metagenome assembly § Result outputs § Others public pipelines

http: //www. ebi. ac. uk/metagenomics