bootstrap results about salmon HOT 10 CLOSED

Hi @s6juncheng,

What do you mean "they do not look at the results"? Because bootstrap values can be large, they are stored in a compressed binary format. Can you tell me what files are in your aux/bootstrap folder? If you are interested in observing the bootstrap values manually, I can help you with a script that converts them to a text format.

--Rob

from salmon.

Hi @rob-p ,
sorry for the typo. Because the file bootstraps.gz is binary, and I saw from http://robpatro.com/blog/?p=291 that, there should be a text file named quant_bootstraps.sf.

Yes, if you have a script for conversion, I would like to have.

I used Kalisto on my data before, but because kalisto do not support strand specific data, I switched to salmon. Does strand specific information helps a lot?

Thanks,
Jun

from salmon.

Hi @s6juncheng,

Ahh, yes, we've since updated the data format so that output sizes stay small even when the number of bootstraps grows large. Thanks for pointing out the inconsistency with respect to the blog; I'll update that. I'll put together a script to convert the bootstrap values to a text format so that you can view them.

Regarding strand specific information: Yes, if you have considerably antisense transcription, then the strand information can be quite helpful. It can make a significant difference in terms of quantification results.

--Rob

from salmon.

Hi @s6juncheng,

I've written this script which will convert the binary bootstrap samples into a text format. The text format is as follows:

• The first line is a tab-separated list of transcript names.
• Each subsequent line is a tab-separated bootstrap sample.
  • The number of entries in each of these lines is the same as the number of transcripts
  • The number of such lines is equal to the number of boot strap samples.

Thus, you can view this as a column-oriented file, where each column contains all of the bootstrap samples for a single transcript.

from salmon.

I've also updated the blog post to refer to the new format. I'm going to close this issue for now, but please feel free to comment on it or re-open if you run into any more trouble.

from salmon.

Hello @rob-p

I tried running your script for converting binary bootstrap samples into a text format. I gave back following error:

Traceback (most recent call last):
File "./ConvertBootstrapsToTSV.py", line 98, in
main(args)
File "./ConvertBootstrapsToTSV.py", line 45, in main
txpNames = nf.read().strip().split('\t')
TypeError: a bytes-like object is required, not 'str'

I processed the samples with salmon version 0.8.2. Do you have any suggestion how I can fix this error?

Thank you.

from salmon.

Hi @tmmse,

Interesting. It should work out of the box. Could you send / share one of your quant directories as well?

from salmon.

Hello @rob-p

I compressed a directory from a sample that was processed using the bootstrap options and added it as attachment.
sample1_bootstraps30.tar.gz

from salmon.

Hi @tmmse,

I have a feeling this is a python 2 vs. python 3 issue. I just ran your data through the script without problem on my end, but my default python install is still 2.7. When using a Python 3 install, I get the same error you're seeing. I updated the script so that it should now be python3 compatible. You can find that version here. Please let me know if it works for you. These changes should be in the next release as well (which should land soon).

from salmon.

Hello @rob-p

I was using python 3.

Your new script works without producing an error. When I run it, I get a text where the first row are the transcripts and the otherline are bootstrapvalues. Thank you very much for the help!

from salmon.