upscaling ChIP-seq data

Question

[R]

Hi,

Could someone please tell me how to upscale a ChIP-seq data?

For down-scaling I use samtools view -bs but in the case of up-scaling I could not find anything

Thanks

[Sean Davis]

What is "upscale"? In other words, what do you want to do?

[R]

to scale unequal datasets, one way is to upscale/downscale the smaller/larger dataset. I want to upscale one of my dataset to increase the number of reads.

[Sean Davis]

You cannot validly do that for any use case that I know of. You could subset the larger one, though.

[Devon Ryan]

Given the number of peak callers out there already, why are you rolling your own?

[Devon Ryan]

Also, have you read the source code for MACS or another peak caller? Many of them allow this and you could just see directly how they do it.

[R]

But I do not want to call peaks! I want to make the tracks with equal reads

[Devon Ryan]

Then subset the larger one.

Answer 1

[bede.portz]

Homer may do what you need. If by "track" you are referring to a file you can visualize on the UCSC browser, IGV, etc, than you can take your reads and use homer to make a tag directory, then use homer to make a bedgraph. Homer defaults to 10million reads for each file, or "track," using the makeUCSCfile command.You can tell homer what number of reads to normalize each dataset to, using the -norm <#> option.

Starting with you reads:

makeTagDirectory <directory> <reads.bed>
makeUCSCfile <directory> -o auto -norm <#>

The first command will create a directory of the reads you want to analyze, separated by chromosome, accelerating subsequent operations. The second command makes a bedgraph, normalized to the number of reads you specify, or 10 million by default. The bedgraph will be placed in the directory used to generate the file (as specified by the -o auto option)

If you want to normalize total read number among datasets for other analyses, such as composite plots or heatmaps, you can do so using homer and the annotatePeaks tool.