Moore Notes 2 9 11

Group Call

• James: SAR for niche mapping project
  • Plots with different average numbers of reads per cell: plot1(0.4), plot2 (4), plot3 (40), plot4 (400)
  • At very high sampling depth, all areas will have all genera
    • This could be real biologically since bacteria genera are big and we're not looking at particularly niche driven taxa (like hyperthermophiles)
    • Filtering out things that only occur in <10 samples
  • There is a linear pattern on the log-log scale at shallower depths
  • Is the non-zero probability output by GLM a realistic framework for this?
    • They are driving the flattening
    • The minimum p-hat depends on the number of samples/reads going into the model fitting
• Should we look at model mis-specification and the dependence of results on sampling depth in general?
  • Can probably just acknowledge limitations for now
• Josh: additional niche mapping updates
  • AUC>0.7 as a criteria for including taxa omits 6 common genera (including Pelagibacter and Prochlorococcus)
  • Draft manuscript is being edited, will send out to group soon (1 week turnaround OK for every one)
• Alex: working on phylum specific richness maps
  • Need to adjust for numbers of genera in each phylum
  • Waiting for new AUC values from Josh