GC-Bias Correction Software

Question

[john]

Hi, my question is basically the same as : GC-BIAS correction for BAM file

I need a software that will take in bam/fastq and output a GC (and other) -bias corrected bam/fastq. The issue with computeGCBias/correctGCBias is that one of their dependencies requires centOS7 to run but I only have access to centOS6.

Are there any other software that corrects for GC bias and outputs a bam/fastq file? I looked at several software like salmon and some R packages but they don't seem to have an option to output bam/fastq.

Thanks for the assistance

[dariober]

one of their dependencies requires centOS7

Can you be more specific? Which dependencies? I installed deepTools 2.1 on CentOS 6 without problems.

[john]

Yeah i installed deepTools no problem but when I ran computeGCBias, I get this error: ImportError: /lib64/libc.so.6: version `GLIBC_2.14' not found

This comment "CentOS 6 uses glibc 2.12 so those errors are to be expected if you try to run something compiled for a more recent glibc version. The solution is either to get the correct application built for use on glibc 2.12 or to use CentOS 7 which has glibc 2.17." from https://www.centos.org/forums/viewtopic.php?t=57881 implies that I cannot install glibc 2.14 on CentOS 6

[Devon Ryan]

This will only happen if you compiled something under centOS7 and are using it then on a CentOS6 machine. The source code itself will compile against whatever version of libraries you happen to have, that won't matter.

Perhaps you'll have better luck with using conda to install things, since this suggests that there are larger issues with your server setup.

[john]

Devon, thanks,

I followed your advice and used conda to install deeptools, and computeGCBias now runs smoothly

Answer 1

[Devon Ryan]

computeGCBias doesn't have any CentOS-version specific requirements. For installation with conda, in fact, the build uses CentOS6.

BTW, one question becomes whether you really need to do this in the BAM files or if you can just adjust the counts (e.g., conditional quantile normalization in the CQN package in R) during the downstream analysis.

[john]

please refer to my reply in the original thread for clarification on the CentOS 6 Issue. I used pip install deeptools but I don't think that changes much.

As for why I want the bam file, it is so that I can visualize the expression profiles before and after GC-bias correction

[John]

Outputting a BAM file is going to be tricky. I don't think deeptools works this way, and even if it did i'm not sure it should.

You can't/shouldn't correct the reads in a BAM file. That would involve making up data, like read orientation, insert length, and a bunch of other things which aren't strictly true. However, you can turn your BAM file into a signal track, by piling up reads, and that signal track can be corrected to give more weighting to certain areas depending on GC content.

So i mean, I think you're going to end up comparing two bigwig files rather than two bam files at the end of all this.

[Devon Ryan]

correctGCBias outputs a BAM file with reads added as appropriate.

[John]

shows how much i know :P

[WouterDeCoster]

So 'synthetic' reads are created by adding more of the same (known reads) to balance the GC bias?

[Devon Ryan]

Correct. I should note that it's rare for any modern (made in the last few years) datasets to need this.

[john]

I'm aware that recent sequencing protocol has improved to account for errors and biases but wouldn't you still want to run various bias/error correction software regardless of when the dataset is made, just in case?

[Devon Ryan]

Nah, wasting time and space is worse. There are always trade-offs to every normalization that you do. Doing normlizations when they're not needed is a nice way to get wrong results (e.g., due to odd side-effects).

[john]

Yeah, I did not explicitly mention it but the intention is to get the bam so I can convert to bigwig to visualize. To do that I need a bam file which is why I was trying to find a program that gives me a gc-bias corrected bam file.

EDIT, I can also just igv the two bam files and look at the coverage too

Answer 2

[Ron]

Check out these softwares :

https://en.wikipedia.org/wiki/List_of_RNA-Seq_bioinformatics_tools#Bias_Correction

http://cnvkit.readthedocs.io/en/stable/quickstart.html