UniFrac: Comparing Microbial

Communities

Presented by:

Donovan Parks

Overview

Motivating example

Quick look at Bray-Curtis Index (BCI)

UniFrac: distance measure, significance test, community tree, sequence jackknifing

Contrast BCI and UniFrac

Why compare microbial communities?

Soil suppressing Tree Pathogens

Soil not suppressing Tree Pathogens

Is this guy protecting our trees?

Bray-Curtis Index (BCI): Distance Measure

Measures the difference in OTU diversity between microbial communities.

Community 1

Community 2

Cluster

ClusterDifferences

Clusters defined by taxonomy or sequence similarity

BCI = 0.4 / 2 = 0.2

Cluster

Normalize

Normalize

0.3

0.3

0.4

0.4

0.4

0.2

0.1

0.1

0.2

Normalize so measure is between [0, 1]

BCI: Pros and Cons

Pros: Simple (both conceptually and experimentally)

Considers both richness and evenness

# OTU in a Community

Relative Abundance

Cons: All clusters considered equally different

UniFrac: Distance Measure

Measures the evolutionary distance between microbial communities

Similar Communities Maximally Different Communities

UniFrac Distance Measure = (------) / (------ + ------)

UniFrac: Significance Test

...#

Tre

es

UniFrac Distance

Magic level of significance!

Randomize Community Labels

Observed Tree

Do two communities differ significantly?

UniFrac: Community Distance Matrix

Pairwise Comparison

0

0

0

0.6 0.5

1.00.6

1.00.5

UPGMA

Distance Matrix Community Tree

UniFrac: Sequence Jackknifing

How does the number and evenness of sequences from each community affect the results?

Sequence Jackknifing

UniFracAnalysis

UniFracAnalysis

Replicate 100 to 1000 times

Weighted UniFrac

What about the relative abundance of sequences (i.e., evenness)?

Since there are twice as many -- sequences, they get half the weight.

UniFrac: A Practical Look

Multiple Sequence Alignment

(e.g., MUSCLE)

16S Gene Sequence

Library

Tree Construction(e.g., RAxML)

UniFrac Analysis16S Gene Sequence

Library

16S Gene Sequence

Library

Filter Sequences

Filter Sequences

Filter Sequences

Tree must be rooted

Tree must have meaningful branch lengths

Remove duplicate sequences if not doing weighted UniFrac.

Community Tree

0.7

0.8

Final Remarks: BCI vs. UniFrac

They are different, equally valid measures!

BCI: taxon-based method which considers richness and evenness

UniFrac: phylogenetic-based method which considers evolutionary divergence

Which to prefer depends on the questions you are trying to answer

Final Remarks: Many Ways to Compare

Many richness and evenness measures (e.g., Jaccard, Sörensen)

Other phylogenetic distance measures (e.g., TreeClimber)

Many interesting and meaningful ways to compare microbial communities:

Richness and evenness Evolutionary divergence Functional diversity

References

UniFrac: Lozupone C, Knight R, UniFrac: a New Phylogenetic Method for Comparing Microbial Communities, Appl. Environ. Microbiol., Vol. 71, December 2005.

Lozupone C, Hamady M, Knight R, UniFrac - An Online Tool for Comparing Microbial Community Diversity in a Phylogenetic Context, BMC Bioinformatics, Vol. 7, August 2006.

Online Tool: http://bmf2.colorado.edu/unifrac/index.psp (excellent help page)

BCI: Bray R, Curtis T, An Ordination of the Upland Forest Communities of Southern Wisconsin, Ecological Monographs, Vol. 27, 1957.

TreeClimber (source of soil sample example): Schloss PD, Handelsman J, Introducing TreeClimber, a test to compare microbial community structures, Applied and environmental microbiology, Vol. 72, No. 4., April 2006.