ATACSEQ normlization for bigwig files

Question

[RD]

Hello everyone! I’m working on an ATAC-seq project with four samples (no replicates), each sequenced to a different depth (e.g. 200 M, 300 M, 400 M paired-end reads, etc.). After marking duplicates on the BAMs, I’m generating BigWig tracks with deeptools bamCoverage using RPGC normalization (as recommended here: https://groups.google.com/g/deeptools/c/th96gaftAXQ). When I run computeMatrix and plot TSS enrichment (cluster of 5 genes), I still see fluctuations instead of a smooth curve that I haven’t been able to explain.

Could someone advise on:

• Whether RPGC is the best normalization strategy when you have no replicates but varying library sizes.
• How to calculate and apply the correct scale factors (e.g. using the --scaleFactor option) if RPGC alone isn’t sufficient.

Any tips on achieving truly comparable BigWig tracks (and hence TSS plots) across samples would be hugely appreciated!

Thank you!

[ATpoint]

e.g. 200 M, 300 M, 400 M paired-end reads, etc.

For the next experiment, consider spending money on replicates rather than 10-fold excessive depth. We typically do 30mio reads per sample. At >> 100mio most will just be duplicates.

Answer 1

[ATpoint]

See for code and discussion ATAC-seq sample normalization

[RD]

Thank you! ATpoint ,Can I use multiBamSummary to get the size factors, then take the inverse (1/sizeFactor) to use as scaling factors in bamCoverage and generate the BigWig files?

[ATpoint]

I do not know this tool. My best practice is in the linked answer.