changing uclust parameters during tax assignment

[Colin Brislawn]

Good afternoon,

I have run the script parallel_assign_taxonomy_uclust.py and want to change the settings that uclust is passed.

One of the lines of my output file reads:

# uclust --input uclust_assign_tax/UCTA_7A16_/UCTA_7A16_.0.fasta --id 0.9 --rev --maxaccepts 3 --allhits --libonly --lib /people/bris469/bin/anaconda/envs/qiime-git/lib/python2.7/site-packages/qiime_default_reference/gg_13_8_otus/rep_set/97_otus.fasta --uc /people/bris469/.qiime_tmp/UclustConsensusTaxonAssigner_B6xMr6.uc

I understand that I can change the --maxaccepts option using the qiime parameter --uclust_max_accepts.

How do I add additional parameters to uclust, like --nowordcountreject.

Thanks!

Colin

[Jenya Kopylov]

Hi Colin,

Not sure it's possible to pass other parameters (that are not accessible via parallel_assign_taxonomy_uclust.py parameters) without modifying the code. I also don't see the option --nowordcountreject in the uclust API (see here), which should be added for building a command.

Perhaps others can add on if I'm missing something?

Jenya

[Colin Brislawn]

Hello Jenya,

Thanks for helping me with this. It looks like the API does not include all the flags listed in the old uclust manual:

http://www.drive5.com/uclust/uclust_userguide_1_1_579.html

The good news is, that API does include the flag --exact, which does what I want! How do I pass --exact to uclust when running parallel_assign_taxonomy_uclust.py?

Colin

[Jenya Kopylov]

Hi Colin,

Seems that the only options you can pass to uclust are those which are supported by parallel_assign_taxonomy_uclust.py. You can create an issue to support additional parameters if this is something you really want to see. Sorry can't be of more help ..

[Colin Brislawn]

Oh that's too bad. 

Maybe Kyle or Greg comment on the possibility of adding this feature. I was hoping it would not be too hard as --exact is already provided in bfillings.

Colin

[Colin Brislawn]

Hey, could I just add this into the command here?

https://github.com/biocore/qiime/blob/master/qiime/assign_taxonomy.py#L1288-L1292

Colin

[Jenya Kopylov]

It shouldn't be too difficult to add. If you want to be able to pass the option via command line,

you'd want to modify /scripts/assign_taxonomy.py as well. Also, I'm not sure whether what

implications this will have on the algorithm, or parsing the output file, something to look into

and test to make sure you're obtaining correct results.

[Colin Brislawn]

It will only change how intensively uclust searches for hits. The output will be the same and the interpretation will be almost identical.

I think I got it working (because uclust is now taking forever). Thanks for your help with this. 

Colin

 

[Colin Brislawn]

Ok, I got it working.

In these lines of this file, you can add it like this:

        # Configure for consensus taxonomy assignment

        app.Parameters['--rev'].on()

        app.Parameters['--lib'].on(self.Params['reference_sequences_fp'])

        app.Parameters['--libonly'].on()

        app.Parameters['--allhits'].on()

        app.Parameters['--exact'].on()

The issue is that this take FOREVER. Setting --maxrejects 0 does NOT search everything; it searches everything that passed the word filter. Passing --exact or  --maxrejects 0 --nowordfilter will force uclust to align a divergent read with literally every item in the database, which is absurd. This seemed more reasonable before I tried it.

Thank you all for your help!

Colin

[Jenya Kopylov]

Glad you got it working!