Question: Correct GC bias using read counts and Loess regression method
0
gravatar for fatima.m.zare
2.4 years ago by
fatima.m.zare20 wrote:

I read this post ( C: Compute Binned Gc-Normalized Read Counts From Bam File ) and I am wondering how can I remove GC content bias with Loess regression?

ADD COMMENTlink modified 3 months ago by Leon90 • written 2.4 years ago by fatima.m.zare20

If you have an issue with GC-bias, why not use computeGCBias and correctGCBias from deepTools?

ADD REPLYlink written 2.4 years ago by Devon Ryan90k

Actually I want write my own code. I want to know about the method that use loess output and do correction for read counts. but Thanks for your answer. I will look at them maybe I can find their methods.

ADD REPLYlink written 2.4 years ago by fatima.m.zare20
2
gravatar for Leon
3 months ago by
Leon90
Leon90 wrote:

Get 300kb bin window:

bedtools makewindows -g hg38.chrom.size -w 300000 > hg38_300.bed

hg38.chrom.size like this:

chr13 114364328
chr18 80373285
chr21 46709983

Get gc content

bedtools nuc -fi hs38DH.fa -bed hg38_300.bed | cut -f 1-3,5 > 300.gc.bed

Get depth in each bin:

bedtools coverage -a hg38_300.bed -b S1901020.aln.bam > S1901020.counts

combine gc content and depth:

paste <(grep -v '#' 300.gc.bed) <(cut -f4 S1901020.counts)|sed '1i chr\tstart\tend\tGC\tRC' > nipt

Rscript

RC_DT<- read.table('nipt',sep='\t',head=TRUE) gcCount.loess <- loess(RC~GC,data=RC_DT,control = loess.control(surface = "direct"),degree=2) predictions1<- predict (gcCount.loess,RC_DT$GC)
resi <- RC_DT$RC-predictions1
RC_DT$RC <- resi
the corrected RC in RC_DT$RC, more help about loess can be got in help document in R.

ADD COMMENTlink modified 3 months ago • written 3 months ago by Leon90
1

for MacOS users, use gsed instead if sed, otherwise you will get an error sed: 1: "1i chr\tstart\tend\tGC\tRC": command i expects \ followed by tex

ADD REPLYlink written 5 weeks ago by Grinch60
Please log in to add an answer.

Help
Access

Use of this site constitutes acceptance of our User Agreement and Privacy Policy.
Powered by Biostar version 2.3.0
Traffic: 939 users visited in the last hour