Moore Notes 5 17 11

Group Call

• Open access

• Protein families in GOS
  • Sam: protein family phylogenetics simulation project
    • read length, reference Db, phylogenetic method
    • rpoB, lolC, 16S - rpoB and 16S are very similar
    • unifrac ROC curves on enterotypes (weighted more sensitive to tree error)
    • could run more proteins, e.g., one or two from each of Steve's clusters
  • Steve: built trees and estimated taxonomic composition for each AMPHORA gene (plus 16S) separately
    • compared patterns of PD across genes
    • different genes give different PD patterns across GOS samples - clustered into 6 groups
    • correlations with environment are different for the 6 groups
    • JE: are all of these still more similar to each other than other genes?
      • KP: scale would change, but negative correlation
    • JE: compare unifrac dissimilarity to other beta-diversity measures to make sure it isn't an issue unifrac per se
    • JG: how does relative abundance play into results? May depend on gene length
    • MEGAN (whole metagenome + pplacer) vs. AMPHORA approach - some taxa are in terms of very different in terms of presence/absence and 5-10% different in terms of relative abundance
      • should look at MEGAN with a subset of genes to eliminate this variable (vs. methods)
  • Overlap with AMPHORA2 paper
  • Syngergy between Sam and Steve's approaches
    • Sam will test if she can get more gene families running, plus pplacer in simulations
    • Can we quantify which gene families are better for phylogeny based analyses