The difference on wiggle file from SICER and MACS

[tujchl]

HI:

I compare wiggle file generated from SICER and MACS (both are 10bp span), and visualized them on IGV. To my suprise, wiggle file from MACS is much better than from SICER on VISUALIZATION. but according to both manual, they all just shift reads and record reads number in each 10bp bin.

some one can tell me the difference?

PS: wiggle from MACS is typically much bigger than from SICER

[ETHANol]

SICER is probably filtering duplicate reads, which for ChIP-seq is often not the best thing to do.

[tujchl]

Thanks ETHANol:
I try to generate wiggle file by SICER without filter duplicate reads. but difference still exists on visualization. I suppose MACS do some kinds of smoothing?

[ETHANol]

First, from my experience with SICER is that it does not create wig files, it creates BedGraph files (similar but not the same). If my memory is correct the BedGraph file comes in the output of running SICER. SICER filters duplicate reads before the analysis, perhaps you can turn this off and you did, but I don't know. If you ran the default, the duplicate reads were probably filtered. But this comes to a more important point, no one can really know because you have not described what exactly you did. So if you describe in detail and include your code, someone may be able to figure this out. Until then we are just wasting time.

[tujchl]

Hi ETHANol:

thank you for your replying

SICER do generate wiggle file and you can refer readme.pdf from SICER V1.1 and below is one part of its readme mentioned wiggle file:
******************
In addition, the two wig files shall be used for visual examination of the raw and processed data on the genome browser.
******************
and I just use SICER sh file to do my analysis:
****************
sh SICER-rb.sh SICER_V1.1/SICER input file . hg18 1 200 150 0.74 400 100
****************
just the same as its example
and wiggle file from SICER is just filter duplicate reads and normalize by total reads

I use MACS (1.4.1) to do analysis:
****************
macs14 -t inputfile -f SAM -g hs -p 1E-5 -m 10,30 --shiftsize=125 -S --space=10 -s 80 --slocal=5000 --llocal=10000 --bw=250
****************

and above all I just want to make sure if MACS do some kinds of smoothing just like Quest, for its wiggle file on IGV looks more smooth than SICER using the same span (10bp) in wiggle file

[ETHANol]

Looks like I'm running a previous version of SICER. Time to update. So that explains some of the confusion.

From what I read SICER: 1)removes duplicate reads from your bed file, 2) makes a bedGraph file from bed file with the duplicates removed, 3) converts the bedGraph file into a wig file normalized by library size.

MACS just takes you bed file and converts that directly into a wig file without the intermediate step where you will lose some resolution (and why it looks smoother). You are keeping the duplicate reads here as well, which for ChIP-seq is often better.

That's my guess of what is going on.

[tujchl]

Thank you ETHANol:

I try to generate wiggle by SICER without filter out duplicate reads.
but difference still exists. MACS looks still smooother

[ETHANol]

MACS looks still smooother

I think that is because you are losing resolution from the conversion from Bed to BedGraph to Wig. It's like a copy of a copy of a copy.

MACS just does a single conversion from Bed to Wig.