Moore Notes 8 24 11

Group Call - Protein Db project

• Dongying ran all new genomes vs. all HMMs
  • Guillaume is parsing the data (takes 6 hours)
  • Found a bug in his code, but fixed now. Re-running today.

• Guillaume working on annotation
  • Distance matrix is too huge for clustering families based on shared Pfams
  • Fuzzy clustering or network based on shared Pfams should be good enough
  • Pfam to GO versus InterPro (better? slow, but should be OK with one sequence per family)
    • Use consensus sequence for each family for annotation mapping
  • Goal: establish a rough guess at number of distinct functional groups

• Tom ran GOS translated reads (~6 million) vs. all HMMs
  • About 75% are classified into a family
  • Next: compare to Rusch et al.
    • Numbers of families
    • How reads are distributed across the meta-proteome

• Parent-child relationships computed
  • Still need to compute betweenness and other graph summary statistics

• Meet next Tuesday in Davis

• Ready to start writing manuscript
  • Morgan has input on introduction (other protein dbs)
  • Outline methods and results (what are the top results for figures)

• Where will data be hosted?
  • Edhar would work
  • Do we want a domain name? Probably only if we make the user interface later