Moore Notes 10 30 13

Group Call

• Participants: Katie, Jonathan, Tom, Patrick, Guillaume, Stephen, Ladan

• PICRUSTs analysis (Tom and Guillaume)
  • English Channel metagenomes and 16S, both annotated to KEGG Orthology Groups
    • For PICRUSTs used GreenGenes 16S tree
    • JE: What genomes were used? This is under the hood, probably same as PICRUSTs manuscript and not appropriate for marine
  • Plots: http://edhar.genomecenter.ucdavis.edu/~gjospin/picrust_test/picrust_test/
    • Weak correlation using abundance estimates
    • Many families with zero or low abundance in 16S/PICRUSTs but high abundance in shotgun metagenomes
    • Large Hamming distances (>0.7) on presence/absence estimates
  • What is different from Jack Gilbert's analysis?
    • Figfams vs. KEGG
    • Similar but not same algorithm as PICRUSTs

• JE will ping Tara Oceans folks again re: data release

• Stephen: Average genome size estimation project
  • Taxon specific? Possible maybe with taxon specific markers
  • JE: Compare to normalizing to number of reads hitting a single copy protein coding gene (e.g., recA)
    • recA is not one of Dongying's markers used in this analysis, because it is diverse between bacteria and archaea
  • Applications:
    • Protein family abundance estimate normalization
    • Ecological differences in average genome size, e.g., IBD significantly different from healthy in MetaHIT
  • Estimates between different libraries on the same sample don't agree as well as one would hope (different lanes of same prep and different simulation runs do agree)
    • KP: Could this be due to duplicate sequences? Probably not, but can check with fastqc

• Next call: Nov 13