Tweet
I'm trying to look for differential open chromatin between 2 groups with two replicates each. After bowtie2 alignment, Group A has 70-76million paired reads while Group B had 80-95million paired reads.
Peak calling with Genrich using replicate mode called out 20k peaks in Group A, 30k peaks in Group B.
I'm guessing there's more identified peaks in Group B mainly because the two replicates in Group B had more reads to begin with. Because there's more called peaks in Group B than A, it's affecting my differential peak analysis.
I used DiffBind for differential analysis and 95% of the differential peaks were unique to Group B. I had to input both the peak files along with the bam files so shouldn't Diffbind's normalization account for the uneven reads between the groups?
Is there a critical normalization step I should have taken early on? Maybe downsampling the fastq files to equal number of reads? This is my first time doing ATACseq and I haven't been able to find any info on read normalization for atacseq.
Peak calling with Genrich using replicate mode called out 20k peaks in Group A, 30k peaks in Group B.
I'm guessing there's more identified peaks in Group B mainly because the two replicates in Group B had more reads to begin with. Because there's more called peaks in Group B than A, it's affecting my differential peak analysis.
I used DiffBind for differential analysis and 95% of the differential peaks were unique to Group B. I had to input both the peak files along with the bam files so shouldn't Diffbind's normalization account for the uneven reads between the groups?
Is there a critical normalization step I should have taken early on? Maybe downsampling the fastq files to equal number of reads? This is my first time doing ATACseq and I haven't been able to find any info on read normalization for atacseq.