Moore Notes 12 18 13

Group Call

• Participants: Katie, Tom, Stephen, Dongying, Patrick, Josh

• PICRUSTs update: http://edhar.genomecenter.ucdavis.edu/~gjospin/picrust_test/picrust_test/Dec_3_2013/
  • Does collapsing KOs to pathways (via EC identifiers) improve agreement between PICRUSTs and Shotmap?
    • Small improvement in correlation, but similar to KO results
    • Correlation is statistically significant but not high
    • Katie: Check that ECs are all pathways and not individual enzymes
    • Discussed results to Jack Gilbert
    • Stephen: What about presence-absence? Tom: slightly better. Stephen: could check families that are above an abundance threshold in metagenomes.

• Stephen: Looks like the Tara Oceans data is available?
  • Josh will check release dates
  • Tom: Check rules about using data pre-publication (Ft. Lauderdale, permissions etc)
  • Where to put data and do compute?
    • Single repository, with mirrors
    • Distribute compute jobs across systems and users
    • IHG cluster
    • Other pay-as-you-go options: EC2 just got cheaper

• Josh: Tibetan soils
  • Can we predict current bacterial distributions from historical climate data?
    • About 80% of taxa (plus community composition: NMDS1) are better predicted by climate data from 50 years ago than current climate data
    • Similar results were found in North American soils, e.g., Deltaproteobacteria
    • Josh: plants and other dispersal limited taxa in communities could be driving this
      • Stephen: how closely do bacteria really associate with plants?
      • Tom: check what lag is in plant data
      • Josh: Will also look at deserts vs. other regions with more plants
      • Stephen: What was happening in 1970s when prediction falls? Is it real or an artifact of data collection?
        • Check what climate variable(s) change then
        • Tom: did soil disturbance patterns change?
      • Tom: Fluctuations are also interesting