Beiko Deep Genomics presentation - "Grand theft operon - lateral city"

http://www.nimr.mrc.ac.uk/images/multimedia/news/large/cake-large.jpg

Robert BeikoFaculty of Computer Science*Dalhousie UniversityHalifax, 2 feet of snow last week, CanadaApril 5, 2014

The dream of a Tree of Life

Can a ToL be[correctly][reliably][accurately]inferred?

Woese

“All happy phylogenies are alike; each unhappy phylogeny is unhappy in its own way.”

- Evolution Leo Tolstoy

Creevey et al. Proc. R. Soc. Lond. B (2004)

Early ancestral signal

is probably gone

It getsworse

W. Ford Doolittle, Sci Am (1999)

OMFG it gets even worse

Kunin et al. (2005) Genome Res

make it stop make it stop

Dagan et al. (2008) PNAS

Do not adjust your model

What is the meaning of this??• Signal saturation + tiny branches that happened a

long, long, long time ago• Other unpleasant biases (G+C, rates, etc)• Lateral gene transfer

Finding LGT

en.wikipedia.org

K-m

ers

or

cod

on

usa

ge

Wang et al. (2001) MBE

Phylo

geneti

c dis

cord

ance

Concordance weighted Discordance weighted

Euchlamydispirokaryotes

Extremarchalsobacteriae

Phylogenetics!MAFs, SPRs, LGTs

Chris Whidden+ Norbert Zeh

Building a MAF by edge cutting

Example case: a & c are sisters in the species tree, but not in the gene tree.What can we do to the gene tree?

• Naïve case: O(3kn)• Fancy refinements: O(2.42kn)• Even fancier refinements: O(2kn) (conjectured)

FIXED PARAMETER TRACTABLE –Exponential in the distance between trees, not the number of leaves

Hypotheses about LGTHypotheses about LGT

The Complexity Hypothesis(Jain et al., 1999)• “Informational” proteins have more interactions

with other proteins in the cell, and are therefore less likely to be successfully transferred than, say, metabolic stuff

• Cohen et al. (2011): forget about function, it’s all about the connections with other proteins in the cell

The Selfish Operon Hypothesis:Lawrence and Roth (1996)

• Genes associate in operons because it facilitates transfer of all constituents of a pathway at once

• If the genes were dispersed throughout the genome, then the selective advantage of a pathway could not be propagated via transfer

The Public Goods Hypothesis:McInerney et al. (2011)

• Genes are public goods that can be freely shared and cannot be excluded from being available

• These genes are constantly acquired and integrated into genomes, invalidating the idea of a unifying Tree of Life

Highways of gene sharing:Beiko et al. (2005)

• Gene sharing occurs preferentially between lineages, and successful gene acquisitions often reflect shared ecology

LGT stories

P. aeruginosaP. fluorescensP. lePewtidaP. syringaeP. entomophilaP. stutzeriP. mendocina

(Catherine) Holloway and Beiko, 2010

“Plume”

ProteobacteriaPlanar is plainer, could be pain-er

Beiko, 2011

244 taxa40,631 trees= Bacterial SPR supertree

LGT patterns for Clostridium

Whidden et al., 2014

Cold case – Aquifex aeolicus & friends

(Rob) Eveleigh et al., 2013

LGT in the Wild

Hehemann et al. (2010) Nature

WHY DO SOME GUT BACTERIA HAVE PORPHYRANASES?

OH

NORISON

Smillie et al. (2011) Science

Lachnospiraceae – Gut / mouth enthusiasts

(Conor) Meehan and Beiko (2014) GBE

“Good” strains ..?

“Not so good” strains ..?

Butyrate production – a crucialfunction, subject to LGT

Finding LGT in the microbiome?• Illumina sequencing - aaaaargh!• Mixed samples! [imagine what happens

when you try to assemble!]• Strain-level differentiation!• etc

What does it all mean?LGT seriously undermines the recovery (and validity?) of the Tree of Life

Even so, aggregation methods (supertrees, etc.) can provide a useful scaffold for inferring LGT events

LGT serves as a useful starting point for hypotheses of habitat adaptation / invasion

Metagenomic data offer new context to LGT events (and genomic data show we should be looking at communities), but present huge challenges to inference

FIN