Transcript MEGAN analysis of metagenomic data

MEGAN analysis of metagenomic
data
Daniel H. Huson, Alexander F. Auch, Ji Qi, et al.
Genome Res. 2007
Early metagenomic

Known phylogenetic markers and subsequent sequencing of
clones


Analysis of paired-end reads
Complete sequences of environmental fosmid and BAC clones


Environmental assemblies


Rough annotation of the metabolic capacity
Distinguish between discrete species and population of closely related
biotypes
Problem of using proven phylogenetic markers(ribosomal
genes, coding sequences)

Slow-evolving genes : distinguishing between species at large
evolutionary distances
What is MEGAN?




Metagenome Analyzer (MEGAN)
Free software.
Deviates from the analytical pattern of previous
Built on the statistical analysis of comparing random sequence
intervals with unspecified phylogenetic properties against
databases



Providing filter to adjust the level of stringency later to an
appropriate level
Laptop analysis


Depends on the related sequences in the databases
Comparing result (BLAST)-> laptop (MEGAN)
Graphical and statistical output
Pipeline




Compare against databases : BLAST
Compute, explore taxonomical content : NCBI taxonomy
Lowest common ancestor (LCA) algorithm
Data sets(Sargasso Sea, mammoth bone, Short E. coli K12 & B.
bacteriovorus HD100)
What we can do with
MEGAN




Species and strain identification
through species-specific genes
Searching species or taxa by find
tool
Distribution of strains of a species
Underlying sequence alignments
Experiments-1

Sargasso Sea

data set


Sanger sequencing
Sample 1-4 from DDBJ/EMBL/GenBank



BLASTX->NCBI-NR


10000 reads from Sample1
Randomly selected a pooled set of 10000 reads from samples 2-4
1% no hits from sample1, <3% no hits from sample 2-4
Filters



Min-score : bit-score threshold of 100
Top-percent : bit scores lie within 5% of the best score
Min-support : isolated assignments it by one read) discarded
Analysis-Sargasso Sea data

1.66M reads, AVG. 818bp by Sanger sequensing

Species profile of 16 taxonomical groups

Environmental assemblies

By analyzing six specific phylogenetic markers

rRNA, RecA/RadA, HSP70, RpoB, EF-Tu, and Ef-G
Result
• Sample1
•~83% reads were assigned to taxa that
were more speific than the kingdom level
•Majority of (8298) were assigned to
bacterial group
•Sample 2-4
•~59% reads were assigned to taxa that
were more specific than the kingdom level
•Majority of (5709) were assigned to
bacterial group
•Alphaproteobacteria, Gammaproteobacteria
by a factor of 2-4 over the remaining 14
taxonomic groups
•Eukaryotes & Viruses : size filtering
•Archaea : May be there is 10times as much
vacterial sequence information in the public
databases
•MEGAN vs. previous (Venter et al. 2004)
•Specific assignment information : LCA
Result-cont.
•Averaged weighted percentage of the siz phylogenetic markers for each
of the 16 taxonomic groups
•Easily detect sampling bias between sample1 and pooled sample 2-4
Experiments-2

Mammoth bone

Data set






Roche GS20 sequencing (Sequencing-by-synthesis)
Sample from 1g of mammoth bone , 28000 years
~300,000 reads, 95bp
BLASTZ-genome sequences (elephant, human, dog)
45.4% of the reads mammoth DNA, others are environmental
organisms (bacteria, fungi, amoeba, nematodes)
BLASTX–NCBI-NR for environmental sequences

Filters : bit-score threshold 30, discard isolated assignment (filtered
2086 reads)
Result


19841 reads to Eukaryota, of
which 7969 to
Gnathostomata
16972 : Bacteria, 761: Archea,
152 : Viruses
Experiment 3

Identifying species from various lead length

Short E. coli K12 & B. bacteriovorus HD100 simulation



5000 random shotgun reads
BLASTX-NCBI-NR
Filters




Bit-score threshold 35
20% of the best hit
Discarded isolated assignments
Result : no false-positive assignment, short read can be used for
metagenomic analysis, albeit at the cost of a high rate of underprediction
Experiment 3-cont.

Roche GS20 sequencing

Data set




2000 reads from random positions in the E.coli K12
~100 bp
BALSTX – NCBI-NR
Filters




Bit-score threshold 35
20% of the best hit
Discarded isolated assignments
Result
Experiment 3-cont.

Roche GS20 sequencing

Data set





2000 reads from random positions in the B. bacteriovorus HD100
~100 bp
BALSTX – NCBI-NR : A in figure
BLASTX – NCBI-NR without B.bacteriovorus HD100 : B in figure
Filters




Bit-score threshold 35
20% of the best hit
Discarded isolated assignments
Result
MEGAN 3(June, 2009)

Suitable for very large datasets


Interests changed


Advances in the throughput and cost-efficiency of sequencing
technology
From ‘which species present’ to ‘What’s different?’
Features


Visualization technique for multiple database
New statistical method for highlighting the difference in a
pairwise comparison
MEGAN3-cont.


Comparing 6 mouse gut with human gut
Clickable, collapsible.