Introducing 

Prezi AI.

Your new presentation assistant.

Refine, enhance, and tailor your content, source relevant images, and edit visuals quicker than ever before.

Loading…
Transcript

Global variation of human gut metagenomes

1. Ruminococcus

2. Bifidobacterium

3. Peptostreptococcus

4. Staphylococcus

5. Lactobacillus

6. Acidaminococcus

7. Fusobacterium

8. Eubacterium

9. Clostridium

10. Coprococcus

11. Escherichia

12. Butyrivibrio

13. Bacteroides

14. Brachyspira

Enterotypes of the human gut microbiome

NATURE

20 April 2011

39 individuals :

(22 faecal metagenomes of individuals from 4 nations) + (published 17 individuals metagenome data from 2 nations)

reads from 39 individuals (6 nations, sanger seq + pyro-seq) => trimming(quality, map to hg18) => assembly & gene prediction by SMASH pipeline

selective pressure from host leads homeostasis, most of species occur in low abundance.

study the presence of abundant functions shared by several low-abundance species could reveal the survival stratigies in human gut.

example : Escherichia's FimA, PapC ; associated with bacterial pilus assembly

map metagenomic reads from 39 samples to 1,511 reference genomes (NCBI, Human Microbiome Project, MetaHIT consortium; extract 16S rRNA gene and assign taxonomy by RDP Classifier)

Trim (lower than 15(phred quality), shorter than 300 bp, and then use makeClip)

Trimming

remove reads mapped to hg18 using BLAT

assemble iteratively using SMASH, predict orf by GeneMark

Assembly &

prediction

1,511 reference genomes are used(NCBI, Human Microbiome Project, MetaHIT consortium; extract 16S rRNA gene and assign taxonomy by RDP Classifier)

Phylogenetic

annotation

map reads to reference genome by WU-BLAST(BLASTN), assign taxonomy of top hit (similarity : >65% for phylum, >85% for genus, length : >75bp & >80% of the read length)

transfer paired-end reads taxonomy to fragment taxonomy => count fragment of reference genome => normalized fragment count by dividing genome size (unassigned fragments was normalized by average genome size )

Abundant functions from low-abundance microbes

twin sample : classify 1,119,519 reads from 154 individuals using RDP Classifier (>= 200bp, >=0.5 confidence score), normalize # of read by average 16S gene copy number in genomes belonging to each genus

Phylogenetic analysis of

external database

danish sample : quality trim & filter 85 danish individuals using FASTX toolkit and map by SOAP

estimate abundance of each predicted gene as below equation (g: gene, r : read, thus gene on a singleton read have an abundance 1)

assign orthologous group in eggNOG to predicted protein by BLASTP, caculate abundance as below (k: eggNOG reference protein, g :predicted gene)

Functional

annotation

align predicted proteins to proteins from KEGG, assign KO (best hit with >= 1 HSP scoring over 60 bits), caculate abundance of each KO, module, pathway as above

cluster the abundance profiles by PAM(Partitioning around medoids) algorithm(this support any arbitary distance measure)

normalize each abundance(genus, OG)to generate abundance distribution , use Jensen-Shannon divergence(JSD) for clustering (see http://graphy21.blogspot.com/search?q=divergence)

Clustering

Detection of entero types, cross national clusters

assess optimal number of clusters using the Calinski-Harabasz(CH) index, cluster validation by silhouette validation technique and cluster similarity by Rand index R

Variaion between enterotypes

Functional biomarkers for host properties

Learn more about creating dynamic, engaging presentations with Prezi