BLAST parameters for large datasets

[renesh]

I am mapping contig sequences with plant database (size 2 GB). I am using tblastx and blastx search tool. Can anyone suggest what parameter for blastx/tblastx i should use for mapping with database?. I am having about 1 million contig sequences.
parameter in case for e value, length to be covered by query sequence etc.

[westerman]

For e-value it depends on how many false positives you can stand. E=1 means that you would expect 1 of your contigs to map to the database by random chance. And many others to be weakly associated to the database. You may wish to cast such a wide net and then take the time to track down the false positives. Personally I use E of 10 to the minus 6 as a first pass. I may miss some interesting correlations but at the same time I can focus in on what I do obtain since those results are going to be strong.

[twaddlac]

The culling_limit and max_target_sequences should also affect the speed. culling_limit discards hits that have n number of reads that are better than it. max_target_sequences are the number of hits to retain. I generally set the max_target_sequences > culling_limit but you can tune it to what you want.

Hop this helps!

[flow_science]

why max_target_seqs > culling_limit?

It isn't immediately obvious to me how these two parameters are related. The reason I believe Twaddlac suggested these parameters is because for any given query, you'd like to retrieve at least as many hits as the number of better matches than the alignment in question. This seems like a recursive definition, because when looking across many species most of the hits for a query will not be the best hit, so you'll just end up printing all the hits no matter what...

I'm pretty sure I'm wrong about something here, but cannot rectify my current understanding of how these parameters function:

-max_target_seqs <Integer, >=1>
Maximum number of aligned sequences to keep

Example 1. max_target_sequences = 5
If target1 is the best match for query1, but targets 2,3,4,5 are also good matches - this will retrieve the 5 alignments in a target-centric manner.

-culling_limit <Integer, >=0>
If the query range of a hit is enveloped by that of at least this many
higher-scoring hits, delete the hit

Example 2. culling_limit = 5
Queries (q1, q2, q3, q4, q5) are all good matches for target 1. However, if any of these queries has higher scoring matches to 5+ other targets, then the query->target1 alignment hit will be deleted.

If the query matches X genes better than a given target, therefore, the query-target relationship is deleted, it seems to me this has nothing to do with number of hits you want to retrieve at the end.

Can anyone explain if/how these two parameters are related? Thanks in advance.