Read count for junctions

[Maoting Chen]

Hi,

The read count for each junction can be generated using the .majiq file for each sample after build step. And similar files (SJ.out.tab) containing the read count for junctions can also be generated by STAR. And I notice there is big difference regarding to the read count of the same junction from both files. Why is that?

If I want to filter out LSVs whose read count is too low, which file should I use?

Thanks,

Maoting

[Paul Jewell]

Hello.

Can you specify which fields you are looking at in the majiq filter to compare with the quantifier TSVs?

To filter on read count, the easiest and fastest method at this time would probably be using $ voila tsv --show-read-counts, and then filter the TSV file from there.

Let me know if it helps.

[Chen, Maoting]

Hi,

Thanks for getting back.

The .majiq file created by 'majiq build' can be processed by python-numpy, which generates a file containing the LSV ID, junction coordinates and read count. And I compare this converted file with the SJ.out.tab file from STAR mapping.

Can $ voila tsv --show-read-counts show the read count of one junction for each replicate? Or it just gives the total read count of all replicates? I can't find this option in the MAJIQ website.

Thanks,

Maoting

Paul Jewell <pje...@biociphers.org> 于2022年7月27日周三 11:46写道：

--
You received this message because you are subscribed to the Google Groups "majiq_voila" group.
To unsubscribe from this group and stop receiving emails from it, send an email to majiq_voila...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/majiq_voila/f80da7df-3952-4a64-9b50-f77aaf919e27n%40googlegroups.com.

[Paul Jewell]

Hi Maoting,

By default for each build group --show-read-counts will provide median counts per group. I think if you will need granularity of per-experiment information the only supported way in the current version is to build with each experiment separately.

[Chen, Maoting]

If I remember correctly, 'majiq build' step does generate .majiq files for single experiments although the group is specified. By saying ' build with each experiment separately', do you mean build the experiments without specifying the group?

Also, the input for voila tsv is .voila files generated by psi and deltapsi. For per-experiment information, do you mean I also need to run majiq psi on single experiments as well? 

Maoting

[Paul Jewell]

Yes-ish to both questions.

For individual build, you will still need to specify experiments under a group but it will just look like

[experiments]

exp1=exp1

exp2=exp2

etc.

And then as you state you will have one $ majiq psi ... ; per majiq file. In this case all will be quantified separately so there will not be any grouping / reducing, and voila tsv will show everything in a flat table.

Let me know if it makes sense.

[Maoting Chen]

Hi Paul,

Thanks for your reply.

I tried as you said and it did work. However, there is still stdev_psi for each junction in the voila tsv files. Why is there still non-zero stdev values when I tried to get info for individual sample?

Thanks,

Maoting