Moore Notes 9 30 09

Group Call

• Finding a time that JE can make the call

• Progress Report
  • Put more stuff in Section C
  • Add names to subsections
  • Clean up headers with no text below

• James' paper was accepted at Ecology Letters

• JG met with Louzapone and Knight
  • They like the super-tree approach Steve is using to make trees
  • Full ML vs. adding reads is more or less the same on the HOT/ALOHA data
  • Starting to look at this in simulations
    • Sam: can generate some more simulated data sets with different parameters if needed

• OTU trees
  • How to build trees from the 16S alignments with metagenomic sequences
    • FastTree
    • RaxML
      • straight ML
      • bootstrap to build the tree
    • RaxML for ref tree, then place metagenomic reads on this tree
      • Tom: How much faster? Steve: 5X for HOT/ALOHA, not sure in general
      • How much less reliable?
      • Can constrain the reference tree topology
      • Sam: is it possible to resolve relationships between added reads?
        • Steve: might not work for big data sets
        • Different ways to do it: add 1 at a time, add all together, add 1 at a time and then resolve, add groups and then fix them into the ref tree
  • Include the full genomes or just the metagenomic sequences
  • JG: How many metagenomic reads in GOS samples?
    • Tom: ~15,000 across whole GOS
    • Testing methods with a subset of ~50 reads
  • Reference sequences are getting lower bootstrap support after adding metagenomic reads
    • Might be due to a read being the sister in the tree
    • Need to run simulations to figure out why
    • Martin: look next level up in the tree (i.e. just assign read to a higher taxanomic placement)
  • How to assess node reliability if bootstrap isn't really valid?
    • repeat ML many times and look at distribution of trees/OTU assignments
      • look at mean of summary measure across these trees
      • can evaluate variance
    • Josh: what about jackknife?
    • Katie: try building trees using sliding windows across alignment
      • Sam: like in EST papers
  • JG: Does confidence in phylogenies affect number of OTUs produced by MOTHUR?
    • Could run a bunch of tree through MOTHUR and look at variance (see above)
  • PLAN: preliminary analysis of simulated data by next week

• Do we have compute nodes with >8 cores?
  • Katie and Sam: not at UCSF
  • Srijak: Some at Davis, will let us know how to qlogin
  • Tom will check CAMERA

• Overlap of Tom's and Morgan's interests (novelty projects)
  • Tom: building trees, calculating phylogenetic diversity at gene family/subfamily level
  • Morgan: looking for completely novel gene families, rank/characterize them in terms of membership across species, see how they might work together to annotate them
  • Tom is looking at changes in known families, Morgan is looking for totally new families