tBLASTn parameters and error with the PredictGenes program

[ptreven]

Hi,

I must say you did great work with BG7. I am trying to use BG7 pipeline for annotation of our Lactobacillus gasseri strain (454 sequencing + Sanger gap closures – in large contigs). Firstly I have question about tBLASTn parameters. What are ussual parameters you use for tBLASTn comparisson (evalue, word size, penalties,…)?

Additionally what Extension_threshold and Overlapping_threshold do you recommend to start with, for this organism?

Secondly I encountered strange error with the PredictGenes program.
Here is the printout of the last few rows:

hit = contig00004 length=117373 numreads=6125
Analyzing hsps hit, there are 1
Entering while... hsps.size()=1
Iteration Q047C0 has 0hits
Iteration Q047E4 has 1hits
hit = contig00009 length=78808 numreads=7139
Analyzing hsps hit, there are 1
Entering while... hsps.size()=1
Iterations finished !!! :)
java.lang.NullPointerException
at com.era7.bioinfo.annotation.PredictGenes.main(PredictGenes.java:446)
C:\Users\PTreven\Downloads\BG7\BG7-master\jars>

Interestingly, this error occurred only when I useed XX_sequences_header_fixed.fna for an input. If I used the original XX_sequences.fna the porgram finished successfully. FixFastaHeaderQC.jar reported no problems.

By the way, I am using Windows 7 on Intel Core i7 3,4 GHz, 8 GB of RAM.

Thank you in advance for all the answers!

PTreven

[pablopareja]

Hi preteven,

Thanks for opening the issue.
Could you confirm me where did you get the file BG7.jar from?
I just checked the line where the exception is thrown and, at least in the last version, that line is actually commented out... 😕

Cheers,

Pablo

[ptreven]

Hi Pablo,

I downloaded the BG7.jar from https://github.com/bg7/bg7 by clicking "Download this repository as a zip file" on 20.2.2013.

Thanx!

Primoz

[pablopareja]

OK, another question then, when you say you ran the program FixFastaHeaders, you did that before generating your BLAST XML files, right?

Perhaps this point is not clearly explained in the documentation but you are supposed to execute this program (when needed) as a preprocessing tool for input FASTA files at the very beginning of the process, that's to say, even before launching any BLAST ( @rtobes , @marina-manrique please correct me if I'm wrong with this)

[rtobes]

Yes, the program FixFastaHeaders have to be run before launching BLAST.

[ptreven]

OK, now it works, thanx!
Since the launching the program was written first in the executionsTamplate.xml I assumed that this is the program to run before PredictGenes...sorry :)

Great, one mystery is solved, two more to go ;)
Thanx again!

Primoz

[pablopareja]

No problem ;)
Now that you say it, it's true that the documentation may be confusing in that sense. I will update it as soon as I have some time.

Regarding your other two questions:

_What are ussual parameters you use for tBLASTn comparisson (evalue, word size, penalties,…)?
Additionally what Extension_threshold and Overlapping_threshold do you recommend to start with, for this organism?

@rtobes , @marina-manrique Could you chime in on this?

Cheers,

Pablo

[rtobes]

It depends on the goals of your annotation but you can try with an e-value of 10E-20 and default values for the rest of tBLASTn parameters.
You can choose the Extension_threshold depending on the similarity that you expect with your reference proteins, the technology of sequencing used (different error rates), your preference of gene misprediction (larger than true genes or shorter). You must set Overlapping_threshold depending on the maximun expected size of overlapping fragments in genes in your bacteria and in some sense it also depends on the Extension_threshold. If you set Extension_threshold to a high value probably you need to be tolerant in overlapping and set Overlapping_threshold also high.

You can test the system with some fragment of an available well annotated genome and analyze the results with different values for Extension_threshold and Overlapping_threshold.

Raquel

[ptreven]

Thanx for all the information! Now I just have to do some annotation :)

Bye!

Primoz