Normalising RNA-seq samples from bam files for UCSC Genome browser visualization

Question

[m98]

I have just over 100 files from an RNA-seq experiment and I have working on converting these BAM files to Bigwig files for visualization on UCSC Genome Browser.

I am new to the field of RNA-seq but from what I understand, I need to scale/normalise the BAM files before conversion to bigwig format. There appear to be all sorts of ways and different methods to normalise BAM files. I have found out that bedtools can allow you to scale samples (using the -scale option) using a specific scale factor. I am tempted to use this option as I was already using bedtools in my script doing the converting of BAM files to Bigwig files.

I guess my question is: which method do I use to normalise my samples? How do I decide which scaling factor to use in my bedtools command?

Thanks.

Answer 1

[ATpoint]

The most convenient solution is IMHO deeptools bamCoverage. It offers multiple options for normalization, binning and strand-specificity. Have a look at the docs.

Answer 2

[trausch]

Alfred (disclaimer: my own tool) can create UCSC browser tracks for paired-end RNA-Seq data

alfred tracks -o ucsc.bedGraph.gz input.rna.bam

The resolution parameter (-r) determines the file size (how aggressively coverage values are binned). By default, the method normalizes to ~30 million pairs.

Answer 3

[Ian]

I have tackled this on a small scale by retaining SAM reads used by htseq-count (--samout), and removing those excluded from the final counts, using sed '/XF:Z:$/d;/XF:Z:__/d'.

After normalisation by DESeq2 I use the scaleFactor to scale the SAM > BAM files, using bedtools genomecov (-scale).

Obviously you have the problem of running DESeq2 with many samples, and R isn't very forgiving. However, I found a thread which may be of use to you: https://support.bioconductor.org/p/73399/