1. Metagenome analysis
Natalia Ivanova
MGM Workshop
February 2, 2012

2. 1. Metagenome definitions:
a refresher course

3. Metagenome definitions
Metagenome is a collective genome of microbial community, AKA microbiome (native, enriched, sorted, etc.).
Metagenomic library (or libraries) is constructed from isolated DNA (native, enriched, etc.).
Metagenomic library can be single-end (AKA standard)
or paired-end

4. Metagenome definitions
Single-end (standard) metagenomic library will produce contigs upon assembly (i. e. longer sequences based on overlap between reads)
Any Ns found in contigs correspond to low quality bases
Paired-end metagenomic library will produce scaffolds upon assembly (non-contigous joining of reads based on read pair information)
Ns found in scaffolds correspond either to low quality bases or to gaps of unknown size
ATGCAAAGGCCGCATCCAGCAGGTT
TACGTTTCCGGCGTAGGTCGTCCAA
ATGCAAAGGCCGCATCC
TACGTTTCCGGCGTAGG
AGCAGGTT
NNNNNN
TCGTCCAA

5. Amplified and Unamplified Libraries
Amplified Library
Unamplified Library
Fragmentation (1ug)
Fragmentation (1ug)
Double SPRI
End repair / Phosphorylation
End repair / Phosphorylation
SPRI Clean
Double SPRI
A-tailing with Klenow exo-
A-tailing with Klenow exo-
SPRI Clean
DNA Chip
Heat Inactivation
DNA Chip
Adaptor Ligation
Adaptor Ligation
SPRI Clean
PCR 10-cycle Amplification
SPRI Clean
DNA Chip
SPRI Clean
DNA Chip
qPCR Quantification
qPCR Quantification

6. Metagenome definitions (contd):
Unless the community has very low complexity (i. e. dominated by one or a few clonal populations), assembly at 100% nucleotide identity will be very fragmented.
What to do with k-mer based assemblies?
Use multiple k-mer settings, combine assemblies with an overlap-layout consensus assembler like minimus2 using minimal % identity of 95%. Tradeoff between overlap length and % identity.
overlap = alignment of reads at x% identity

7. Reasoning behind combining multiple assemblies

8. Trimming does not appear to be ideal for this process
Assembly Pipeline v.0.9
CPU time intensive, no known metagenomic Kmer prediction algorithm
A snapshot of older (454-Illumina) metagenome assembly pipeline
Picking best kmer – manual process 8

9. Metagenome definitions (contd):
Assembly of sequences at less than 100% identity =>
population contigs and scaffolds representing a consensus sequence of species population
isolate contig species population contigs
overlap = alignment of reads at x% identity

10. 2 more important definitions
Sequence coverage (AKA read depth)
How many times each base has been sequenced => needs to be considered when calculated protein family abundance
Per-contig average coverage
Per-base coverage => per-gene coverage
2. Bins
Scaffolds, contigs and unassembled reads can be binned into sets of sequences (bins) that likely originated from the same species population or a population from a broader taxonomic lineages

11. What IMG does and doesn’t do
Scaffolds and contigs are generated by assembly – not provided in IMG/M
Sequence coverage can be computed by the assembler based on alignments it generates (preferable) or can be added later by aligning reads to contigs – the latter can be provided in IMG/M
Bins are generated by binning software – not provided in IMG/M
Scaffolds, contigs and unassembled reads are annotated with non-coding RNAs, repeats (CRISPRs), and protein coding genes (CDSs); the latter are assigned to protein families (COGs, Pfams, TIGRfams, KEGG Orthology, EC numbers, internal clusters) – is provided in IMG/M

12. What’s the difference between IMG and MG-RAST, IMG and CAMERA?
We prefer to assemble the data
longer sequences -> better quality of gene prediction and functional annotation
longer sequences -> chromosomal context and binning -> population-level analysis
But we don’t provide assembly services except for metagenomes sequenced at the JGI
we may be able to help with assembly of 454
we’re not equipped to assemble massive amounts of Illumina data
Contact person: Ed Kirton,
IMG does not provide tools for analysis of 16S data from the metagenome itself
we do assembly -> none of assembled 16S sequences is reliable
BLASTn of reads matching conserved regions is misleading
we do pyrotags for every metagenome sequenced at the JGI

13. 2. IMG/M features: divide and conquer
(see also IMG/M -> Using IMG/M -> Using IMG/M -> IMG User Guide and IMG/M Addendum)
username: public
password: public

14. IMG/M User Interface Map About IMG/M -> Using IMG/M -> User Interface Map

15. Dividing the contigs by GC content or length
Statistics
Microbiome Details -> Genome Statistics -> DNA Scaffolds
Search
Microbiome Details -> Scaffold Search

16. Dividing the genes phylogenetically: Phylogenetic Distribution
Phylogenetic Distribution of Genes
Microbiome Details -> Phylogenetic Distribution of Genes
Components:
histograms
Protein Recruitment Plots
summary statistics tables
lists of genes
histogram
(phylum/class)
gene counts
gene lists
summary statistics
(family)
(species)
counts, lists, statistics
counts, lists
recruitment plots

17. Dividing the contigs: Scaffold Cart
Lists of contigs or genes in Gene Cart
E. g. Microbiome Details -> Genome Statistics -> DNA Scaffolds -> scaffold counts
Scaffold Cart
Features:
Scaffold Export
Adding all genes to Gene Cart
Function Profile (against functions in Function Cart)
Histograms by GC content, length and gene count
Phylogenetic Distribution

18. All Carts in IMG are interconnected
Gene Cart
Scaffold Cart
Function Cart

19. Dividing the genes by abundance/ by function
Abundance Profiles
Compare Genomes -> Abundance Profiles Tools
Components:
Common parameters:
Normalization (none/scale for size)
Type of count (raw counts/estimated gene copies)
Type of protein family (COG, Pfam, Enzyme, TIGRfam)

20. Other tools Phylogenetic Marker COGs
Find Functions -> Phylogenetic Marker COGs
SNP BLAST and SNP Vista
Gene Page -> SNP BLAST -> SNP VISTA
IMG/M exercises:
The first 3 pages are questions without answers; the rest is a cheat sheet

21. Life outside IMG: binning tools
Alignment-based tools
MEGAN – BLAST+LCA
MTR – BLAST+ MTR
SOrt-ITEMS – processed BLAST best hit
CARMA and Web-CARMA – MSA + neighbor-joining tree
Compositional tools
PhyloPythia – 6-mers, SVM
TACOA – 2-6 mers, k-nearest neighbor classifier
Phymm and PhymmBL – Interpolated Markov models (IMMs)
ClaMS – DOR, DBC

22. Life outside IMG: statistical analysis tools
Comparison of 2 samples
MEGAN -
STAMP -
Comparison of sets of samples
ShotgunFunctionalizeR – R package for statistical analysis -
METAREP – package from JCVI, includes multidimensional scaling, hierarchical clustering, etc -
METASTATS – package for analysis of paired samples with replicates -
LEfSE – package for comparison of multiple classes of samples with replicates -