Bias Model pearsonR<-0.5 #275
Description
Hi! I have four bulk ATAC samples(2 treatment, 2 control) that I used for training my bias model strictly following instruction from the Wiki tutorial. All four samples have at least 30M reads, with both treatments' FRiP score at 0.2-0.3, and controls' FRiP score at 0.05 and 0.17. All samples have an TSS score>7, meeting the ENCORE suggested cutoff for good quality data. All samples were processed without Tn5 bias correction.
For training in the bias model, I merged the two replicates into one combined treatment and one combined control, and trained the bias model for each of them separately. With a bias factor of 0.5, both samples had non.peak pearson R around -0.3, peak pearson R around -0.6, non peak median jsd around 0.7, and peaks median jsd around 0.6. The learned motifs contain known TFs.
I have tried training the bias model with a bias factor of 0.3, 0.4, 0.6, 0.7, 0.8, but none of these solves the issue. The pearsonR is below -0.5 for all bias models, thus can not be used for training the bias-correction model. I am unsure what might have happen in my sample/training that could cause this issue and would really appreciate any suggestions!
Thanks for spending time reading my message and looking forward for your reply.