Moore Notes 10 20 10

From OpenWetWare
Jump to navigationJump to search

Group Call

  • Steve: PLoS paper has been in review for 2 months - should hear soon.
  • Morgan: Gerstein lab paper uses ARCGIS data - might be useful for niche mapping
  • PhylOTU revision
    • Rebecca's streamlining by putting code into C sped it up a lot!
      • Will share code with every one
      • Dongying can use it for pruning large trees
    • Also found tree pruning bug (removeAncestralNode function in BioPerl)
      • This is called by smoothing function
      • Accuracy improves after fixing the bug
      • Scaled threshold is closer to PID threshold, e.g., 0.03 is now 0.06, not 0.15 at 80% accuracy
      • Threshold of 0.15 is 90% accuracy
    • Will redo simulations this week
      • Initial results show little correlation between average PD of sources and accuracy
      • Using SILVA and/or other simulations with many reads compared to number of reference sequences to make sure we cover a good range of PD (an order of magnitude more PD)
    • Will work with Steve to redo the GOS analysis
  • Niche mapping
    • Talked to Catherine Graham at Stony Brook about bias in environmental layers
      • Pick layers that you expect to drive diversity
      • Layers shouldn't have pairwise coefficient of determination >0.75
      • Appropriate granularity
    • Using these, picked 15 variables from Reddy et al
      • Only the multiple salinity and multiple temperature variables are highly correlated
      • Pick 8 to use
      • Might want to drop coral and estuary (many zero values globally)
    • May want to add: ocean PH, depth (e.g., area in cell <100m), tidal range
    • Could do more than once with different layers
    • What are the criteria for picking relevant variables?
    • Could add a neutral or random layer is a control
    • Estimation versus model selection etc. (after thresholding)
    • Sampling bias layer can be fed into MaxEnt to correct for this bias
      • Might want to account for size of sample: number of bp or number of reads or measure of species coverage
      • Goal is to remove regions of low diversity due to not sampling
    • Try to wrap this up soon: goal to submit before winter break
    • Journal?
      • General biology journal: Nature, Science, PLoS Biology, PNAS