[lucila]

Hi Lucila,

No, it does not, but I have been considering adding gff output, at least for exon boundaries, if not full transcripts. For generating a reference transcriptome, you can do either mapping or assembly; so, assemblers like Trinity are also an option.

I have not tried it, but you may also want to examine this:
https://github.com/enormandeau/gawn

Thank you Brian for the information!!
Cheers,
Lucila.

[Brian Bushnell]

Hi Brian,
thank you so much for this useful tool. I would like to ask you a question. Does BBMap generate a list of transcripts as a result of the mapping of the reads? What I mean is if this tool generates a fasta file of reconstructed transcripts based on the reference genome and the reads used for mapping.
I need this because the genome that I am using is not annotated (I mean, I do not have a gff) and I want to generate a reference transcriptome using my reads and the genome.

Hi Lucila,

No, it does not, but I have been considering adding gff output, at least for exon boundaries, if not full transcripts. For generating a reference transcriptome, you can do either mapping or assembly; so, assemblers like Trinity are also an option.

I have not tried it, but you may also want to examine this:

https://github.com/enormandeau/gawn

[lucila]

Hi Brian,
thank you so much for this useful tool. I would like to ask you a question. Does BBMap generate a list of transcripts as a result of the mapping of the reads? What I mean is if this tool generates a fasta file of reconstructed transcripts based on the reference genome and the reads used for mapping.
I need this because the genome that I am using is not annotated (I mean, I do not have a gff) and I want to generate a reference transcriptome using my reads and the genome.

Thank you again!
Best,
Lucila.

[Brian Bushnell]

Thanks that's helpful. On the note of loading references - is there a way to use wildcards with the input and output of bbwrap?

Ah, sorry, but nope. You can, however, do something like...

ls *.fastq.gz > ls.txt

...then replace '\n' with ',' using a text editor (like Notepad++). I'm sure there's a simpler sed/awk solution, too.

[darthsequencer]

To maximize speed when you are not looking for low-identity matches, "fast" (plus your identity threshold) is generally adequate. You can also speed it up by reducing "maxindel" (fast sets it to 80). Quality-trimming and adapter-trimming generally increase alignment speed.

With a large reference you may be able to increase speed with "k=14" instead of the default "k=13" - this increases the time to load the reference and memory usage, but increases mapping speed (so whether the process becomes faster or slower depends on how long it takes to load the reference compared to how much data you have to map). Also, turning off mate rescue (rescue=f) or reducing rescuedist (fast defaults to rescuedist=800) can also increase the speed slightly. Note that all of these options reduce sensitivity (aside from trimming which increases it), but at 97% identity you only need very low sensitivity anyway.

Thanks that's helpful. On the note of loading references - is there a way to use wildcards with the input and output of bbwrap?

[darthsequencer]

How long are the query sequences?

They range between 50bp single end to 2 x 250bp

[Brian Bushnell]

To maximize speed when you are not looking for low-identity matches, "fast" (plus your identity threshold) is generally adequate. You can also speed it up by reducing "maxindel" (fast sets it to 80). Quality-trimming and adapter-trimming generally increase alignment speed.

With a large reference you may be able to increase speed with "k=14" instead of the default "k=13" - this increases the time to load the reference and memory usage, but increases mapping speed (so whether the process becomes faster or slower depends on how long it takes to load the reference compared to how much data you have to map). Also, turning off mate rescue (rescue=f) or reducing rescuedist (fast defaults to rescuedist=800) can also increase the speed slightly. Note that all of these options reduce sensitivity (aside from trimming which increases it), but at 97% identity you only need very low sensitivity anyway.

[GenoMax]

Hi Brian,
I have a lot of reference sequences I'm mapping to (~11 million)

How long are the query sequences?

[darthsequencer]

bbmap fast macro?

Hi Brian,
I have a lot of reference sequences I'm mapping to (~11 million) and want to eek out as much as speed as possible.

I'm mostly looking for close matches - ex. I set minid to 0.97. Will setting fast still find matches like that? Any other thoughts on what I can set to get more speed?

Thanks a bunch!

[darthsequencer]

It is tied to the number of threads defined for BBMap, just for some reason I capped it at a max of 8 even if the main process was allowed to use more; probably to conserve memory. I've increased it to a max of 64.

Thanks- that helps a lot!

[Brian Bushnell]

How about tying the number to the number of threads specified for BBMap? That way we know that many threads are available.

It is tied to the number of threads defined for BBMap, just for some reason I capped it at a max of 8 even if the main process was allowed to use more; probably to conserve memory. I've increased it to a max of 64.

[GenoMax]

Oh, yep, for some reason I capped it at 8 threads. I wonder why? I'll eliminate that cap in the next release, which will probably be sometime today.

How about tying the number to the number of threads specified for BBMap? That way we know that many threads are available.

[Brian Bushnell]

Oh, yep, for some reason I capped it at 8 threads. I wonder why? I'll eliminate that cap in the next release, which will probably be sometime today.

[darthsequencer]

Hi - I love that bbmap and its tools can directly make bam files. I noticed that it's using samtools with 8 threads. Is there a way to increase the number of threads?

Thanks

[jweger1988]

Thanks for the reply. That is correct.

I have a virus that I introduced some degenerate nucleotides in to track bottlenecks.

I suppose I could just reformat to the area of the read I'm interested in and then just convert to fasta and use that.