Pathway prediction in (meta)genomes usingPathway Signature GenesLucas Brouwers, MSc student, Nijmegen

GoalHow to predict the metabolic capacity in a metagenomic sample, given incomplete data?Current practice:Percentage of sequences assigned to pathways used for estimating pathway abundance2 out of 5 sequences map to pathway X:Abundance of X is 40%We propose to identify the (metabolic) pathways present in a community, making fulluse of the metagenomic data available

Percentage of sequences assigned to pathways used for estimating pathway abundance

ApproachWe use presence of OGs to predict presence of pathwaysSome species have pathway XXXXXOG AOG CXOG BPWY XXPresenceSignatureWeak SignatureAbsenceSignature

Approach630 species and their OGs1,200 (metabolic) pathways

Signature genes add informationDo the signatures add information on the presence of a pathway?(after all some pathways are rare, others are ubiquitous) Specific information defined as a difference in Shannon entropy: how much extra information does the presence of an OG give about the presence of a pathway?

What makes a signature?N = 29,661N = 46,176,543

Absence signatures in a pathwayCOG0437 (221 species): Fe-S-cluster-containing hydrogenasesFormatedehydrogenaseFormylmethanofuran-dehydrogenaseGlycolaldehyde-dehydrogenaseNitrate reductase

Integrate signature scores with pathway prediction in metagenomesFor quantitative analysis of pathways in metagenomic samples:Consider Chlorphyll biosynthesis:Sa: Average of all OG scores, of OGs in species without chlorophyll biosynthesisSp: Average of all OG scores, of OGs in species with …Si: Average of all OG scores found in a metagenomic sampleCorrecting for genome sizes becomes …

Performance in sub-sampling Sample percentage of OGs from every species in STRING Predict percentage of species with pathway & compare with actual occurrence

Sample percentage of OGs from every species in STRING

Predict percentage of species with pathway & compare with actual occurrence

Application to metagenomesDinsdale, E. A., Edwards, R. A., Hall, D., Angly, F., Breitbart, M., Brulc, J. M., et al. (2008). Functional metagenomic profiling of nine biomes. Nature, 452 (7187), 629-32. Different biomes: SubterraneanHypersaline Freshwater Fish Coral Marine

Subterranean

Hypersaline

Freshwater

Fish

Coral

Marine

Simulated metagenomes Sample OGs according to species distribution in metagenome Predict percentage of species with pathway & compare with actual occurrenceSampling 1%

Sample OGs according to species distribution in metagenome

Predict percentage of species with pathway & compare with actual occurrence

Pathway description of biome diversityPathway occurrence was predicted in 35 metagenomic datasetsPrincipal component analysis revealed the pathways responsible for differences between the biomes

ConclusionsPathway signature genes allow us to interpret biomes on a pathway levelGenes do not need to be part of a pathway to predict its presenceWe can quantitatively and accurately describe the pathway content of a metagenome

AcknowledgementsBas E DutilhMartijnHuynen