Do you keep this batch or discard it?

Question

[Parham]

Hi!

I did some quality control on my data with RSeQC. I used geneBody_coverage to check how even is my reads across gene bodies. I am almost sure that one batch should be discarded but I am hesitant about the other one. Below are the images. The batch with sample number from 1 to 6 have heavy bias but I am not sure if the skewness in second batch (7-12) should be considered as bias or not?

[Devon Ryan]

Thanks for the updated links. I guess this depends on what you want to do with the data. If you're doing regular DE analysis then just put the batches into your model. If you're trying to look at splicing differences toward the 3' end, though, you might have to toss at least the first batch.

[Parham]

Thanks for looking into it Devon. My aim is to find DE and then do some downstream analysis like GO enrichment analysis or pathway analysis. Also I am interested in looking into clustering of DEG across genome. If they are present more at specific sites or they are just randomly dispersed across the genome. So you think this library can be used for this purpose? Also here I am attaching some images of heatmap of 50 genes with highest variance across samples and PCA plot for them. Can I have your comments on that too? Would you recommend to continue with these libraries or just drop them?

1. Heatmap and PCA for sample 1-6 in geneBody image:

   

2. Heatmap and PCA for sample 7-12 in geneBody image:

   

[Devon Ryan]

We would need OneDrive logins to see your images. Try figshare or imgur or something like that.

[Parham]

This link doesn't need a log in: http://1drv.ms/1QvYi0l