Moore Notes 11 20 13

Group Call

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

• Patrick: domain specific markers to estimate percentage of eukaryotes in sample
  • Curtis Huttenhower has a database of phylum specific gene families
    • Not sure if they are good for getting relative abundance (i.e., not necessarily even copy number)
  • Dongying is focusing on markers that work for bacteria (not necessarily absent from archaea)
    • For poorly sampled groups, gene presence isn't a particularly good marker for a clade
  • Stephen: what about kmer-based binning?
    • JE: not likely to work
  • Tom: check the microbiome protocols to see if they actually extracted eukaryote DNA
  • Other markers?
    • JE: ncRNAs
    • KP: noncoding DNA
    • Lineage-specific transposable elements

• Tom/Guillaume: PICRUSTs vs. Shotmap
  • Previous analysis was using loose classification parameters that were optimal for shorter reads (RAPsearch score threshold 31)
  • Repeated analysis using score threshold of 100 http://edhar.genomecenter.ucdavis.edu/~gjospin/picrust_test/picrust_test/Nov_18_2013/
    • Still finding many families with shot map that are not detected by PICRUSTs
  • Jack Gilbert's analysis focused on 1050 families predicted by both methods
    • We have ~2000 KOs detected by both methods
    • Our correlation on these KOs is about 0.6 (almost as high as theirs) http://edhar.genomecenter.ucdavis.edu/~gjospin/picrust_test/picrust_test/original_analysis/top100/top100_Rho.txt
    • Could also check if our genome db is different from Jack's and if it is inappropriate for marine
  • Stephen: from simulations, at optimal threshold for relative abundance analysis, quite a few families are predicted as present that are actually absent
    • But we also see differences in relative abundance and presence for more abundant families (according to shotmap)
  • Tom will follow up with Morgan, then Jack

• Josh: soil bacteria niche modeling
  • See slides (emailed to group)
  • Tibet
    • Tom: try unifrac distance (vs. Bray-Curtis)
    • JE: put sampling locations on heat maps
    • JE: assumes that the model between env and bacterial distributions haven't changed
      • Organisms probably haven't evolved too much overall
      • But the organism-organism relationships might have changed: should check if relationships affect the niche model
    • Try past environment with current microbe distributions to see if model is different
      • Might be able to also address if past conditions better predict current conditions
  • N America