Notes 20101118 BIOL 614 Presentations - Greg Vey

From SnOwy - Ed's Wiki Notebook

Metagenome analysis -- Functional annotations

• orthology hindered by loss of discrete genomic units
• IMG/M -- environmental datasets -- got 136 metagenomes
• functional annotation; homology based approached -- COG categories
• COG curated by NCBI
• improve / extend / infer annotation with functional inference -- transitively cascade onto other members
  • physically near genes likely to be functionally similar
• selected likely method: homology based genomic context method ... -- gene neighbour method -- same contig
• experiments
  1. one member has an annotation -- cascade annotation from one member to the other
     • use as benchmark for how good this method is
     • used gene neighbour method -- must be proximal and similar enough (log-likelihood)
       1. about 56% of gene neighbours are subject to transferrence of COGs given both the above
       2. only similar -- 34%
       3. only contiguous -- 10%
  2. how worthwhile is this activity?
     • running this, we get a gain from previous 28% to now 41% new annotated genes
  3. applicability
     • what characteristics are needed in a metagenome that make them compatible with this method?
• the current trend is no longer to find long contiguous genomes
• current / application -- able to apply homology-based annotations