Using Search

1. The search command runs a mixture of:
   • Randomized wagner builds
   • TBR swapping
   • Nixon's ratchet
   • Exhaustive DO
   • Tree fusing

2. search should be the first analysis option for both new and expert users. Let's start with a very small data set, so that we can get some meaningful results with a very small amount of time:

               wipe ()
               read ("18s.fasta")
               search (max_time:0:0:1) 
               (* This will run a search for exactly 1 
               minute *)
   

3. After this command finishes, you should see a message on screen telling you how many trees where built, how many fusing generations where performed, how many times the best tree was found, and what its score is.

4. We can constraint the search some more. Suppose that we have, from previous searches, the impression that the best tree that we could find has cost 385, hen we can tell POY that this should be the target cost for an expected number of hits:

               wipe ()
               read ("18s.fasta")
               search (max_time:0:1:0, hits:5, target_cost:385)
   

   This command will now run for one hour or until it has found 5 trees with cost 385 or less, whichever happens first.

5. If we have limited memory resources, we can now execute this search with a memory constraint, so that POY will only store as many trees as it can fit in 256 MB or RAM, not more.

               read ("18s.fasta")
               search (max_time:0:1:0, hits:5, target_cost:385, 
                   memory:mb:256)
   

6. So lets run a regular search for half an hour on the 18s data set, and see what results you get. We will store the resulting trees in the file 18s.tree:

               read ("18s.fasta")
               search (max_time:0:0:30, memory:mb:256)
               select ()
               report ("search_18s.tree", trees)