Hi everyone,
I apologize if this has already been asked, but I was wondering if the organizers could provide some more detail about how the RNA-seq data in syn21901840 and syn21990044 were processed?
Some additional information would be very helpful so we can make sure we chose the right methods and statistics for differential expression.
Thank you!
Best regards,
Clemens
Created by Clemens Hug chug Yeah, no worries. So nearly all of our QC, benchmarking and analytical work has been done with this procedure so I would be hesitant to deviate too much from the form of the data that we understand the best.
You are correct that invariably, this type of procedure will reduce some information and so I understand your concern. That said, this procedure was optimized by a predecessor of mine and so I wouldn't feel comfortable deviating providing differently normalized data until I could get more information on precisely why this step was added to the normalization procedure. @efd2115 What do you think about this? @allawayr @efd2115 The variance normalization for this data is pretty heavy here. I think there might be a large variance washout that happened somewhere in these steps, which is limiting the prediction quality of the data. I haven't had much luck de-convolving the variance, and was wondering if you'd consider adding the same data without steps 1/2/3 or 2/3? Hi @chug,
This hasn't been asked here but was asked in the previous Panacea Drug Activity Challenge. I'll copy @efd2115 's response here and tag him in case you have additional questions:
>So first of all the PlateSeq method involves 3' sequencing (not whole transcript) and so the effect of transcript size is taken care of by the technology.
>
>From the counts processing stand point the expression normalization procedure is:
>
> 1. Quantile Normalization
> 1. Variance Stabilization
> 1. Batch correction of Plates with ComBat