Biostar Beta. Not for public use.
Calculating correlations using data from different experimental methods for a co-expression network
0
Entering edit mode
14 months ago
pixie@bioinfo ♦ 1.4k

Hello, I have data from RNA-seq as raw read counts from 3 replicates from control and 3 replicates from treatment. I have to calculate Pearson correlations (PCCs) among the genes after using usual normalization procedures. One of the important transcription factor which is important for the biological hypothesis does not have any reads mapped to it, except in one of the replicates.

The biologists want me to use the expression data from RT-PCR experiment for the particular TF for calculating the PCCs, while using raw read counts for other genes. Is this okay to do or is there any solution to this ? Thanks.

RNA-Seq pcr • 187 views
ADD COMMENTlink
2
Entering edit mode

No that doesn't sound okay. What biologists want is not always possible, I have this problem as bioinformatician also all the time.

ADD REPLYlink
1
Entering edit mode

Unfortunately, Its becoming very hard to explain them the logic behind it -_-

ADD REPLYlink
0
Entering edit mode

I am afraid that it is part of the job.

ADD REPLYlink
0
Entering edit mode

Sounds like you have a RNA-seq expression matrix but that it does not include the key gene of interest? So, your colleagues asked you to, instead, use RT-PCR data for this gene of interest?

ADD REPLYlink
0
Entering edit mode

Correct, zero read counts for the given gene, so plug in RT-PCR data

ADD REPLYlink
0
Entering edit mode

I mean, you could just correlate this gene's RT-PCR data to all other genes. As it is correlation, it is not sooo affected by the different distributions. I would use Spearman, though, not Pearson. Your sample n is low, so, non-parametric correlation is a better fit.

I imagine that RT-PCR data is normalised and log2 transformed?

ADD REPLYlink

Login before adding your answer.

Similar Posts
Loading Similar Posts
Powered by the version 2.1