I was wondering how to run a pearson correlation test between two chip-seq data sets. I have narrowpeak files for both which are essentially bed files.
Is there any bioconductor packages in R that can do this? I've tried a variety of bedtools, and programs such as CHANCE, but they haven't given me what I required.
I used MACS2 to call peaks.
I used Bowtie to align to hg19.
Thanks!
2 answers
Try the bamCorrelate function from the deeptools software
•
written
by
Constantine
We have updated to deepTools 2.0 and now bamCorrelate is called multiBamCoverage. The output of that program can be used to compute a correlation (using plotCorrelation) of a PCA (using plotPCA). The new documentation is here:
http://deeptools.readthedocs.org/en/latest/content/tools/multiBamCoverage.html
The "number of uniquely aligned reads per PolII gene" is really just a row of a table of integers.
Sample Gene1 Gene2 Gene3
Rep1 10 20 30
Rep2 12 18 40
For that table, you can compute the correlation in R like
> cor(c(10,20,30),c(12,18,40))
[1] 0.9496528
and say they are 95% correlated.
So your first step is to generate the table, specifying gene column headers to help you count reads at each site.
•
written
by
karl.stamm
There's a lot of different Biostars questions asked and answered for counting reads. Look for "how to count reads". The most popular tool is probably "htseq-count"
I like coverageBed from bedtools. You give it the BAM and a BED annotating your regions of interest, and it spits out how many reads were in each site.
•
written
by
karl.stamm
Log in to answer this question.
-
Best way to combine biological replicate peaks?
written by SpartanII •
-
Different peak sites from .xls and .narrowPeak, macs2 output
written by sh.kazempour94 •
-
IDR on blacklisted region-filtered narrowPeak
written by endicytosis •
-
chip-seq analysis and peak annotation
written by tarek.mohamed
-
peaks overlap: ATAC & CHIP seq
written by sophialovechan •
-
macs2 callpeaks genome
written by sophialovechan •
-
groHMM input data
written by dally
-
DeepTools Error: Be sure that the bam files you are using is indexed.
written by dally
-
Converting .narrowPeak file to GRanges for use in Bioconductor ChIPseeker
written by dally
-
Samtools mpileup refrence fasta file input
written by cvu
Do you know what Pearson correlation is? It measures the collinearity between two one dimensional measures.
I am only vaguely familiar with the test. I've seen it on a large amount of papers recently while reading up on a new project. Essentially what I am trying to do as quoted from a paper is "find the number of uniquely aligned reads per Pol II gene for two complete biological replicates generated from * insert cell type * cells (Pearson's correlation, R = 0.##)
There's no test. There's no hypothesis, and there's no p-value. The correlation is just a measurement.
You can window the data along the genome and calculate a correlation for the values in each window, but you definitely want to filter low windows. This post had some additional ideas.