0
$\begingroup$

I have RNA-seq from two sequencing batches; Lab technician says that he has run the RNA expression quantification two times in bathes 1 and 2 for example tumor 1 in batch 1 and tumor 1 in batch 2 , normal 2 in batch1 and normal 2 in batch 2. PCA shows however shows two experimental runs are too close to each other and can be merged

[![enter image description here][1]][1]

Now my question is can I compare only a tumour sample with its own matched normal sample in one patient to get differentially expressed genes for this patient by using experimental runs (technical replicates )? Design would be Tumour1_batch1 Tumour1_batch2 versus Normal1_batch1 Normal1_bacth2

> head(mycols)
              Condition
Normal1_batch1  N
Normal1_batch2  N     
Tumour1_batch1  T     
Tumour1_batch2  T     
$\endgroup$
  • 1
    $\begingroup$ Yes, you could (although I think you would need three samples in each side to be able to calculate the sd and var). However, how generalizable it would be ? What is your biological question behind this comparison? One thing is to pair the samples and control for batch effect and another is to just compare 4 samples. $\endgroup$ – llrs Sep 17 '19 at 12:54
  • $\begingroup$ My main question here is: what are the differentially expressed genes between tumour and normal adjacent sample for this specific patient where I am going to use two experimental runs as my replications for having statistical power . Thank you $\endgroup$ – Angel Sep 17 '19 at 12:57
  • 4
    $\begingroup$ Using technical replicates tells you how good your technician can use a pipette, it gives you no information about the biology behind it. In statistics we want to sample (take samples from a population) to get info about the whole population. For example, if you want to have the average length from a population, you sample different people to get a good average. In your case you would measure one person twice, it tells you nothing about the population but it tells how good your measuring skills are. $\endgroup$ – benn Sep 17 '19 at 13:48
  • $\begingroup$ If you want to compare individual patients, why are there less normal samples compared to tumor samples? $\endgroup$ – PPK Sep 18 '19 at 13:51
  • $\begingroup$ Because I don't have real replication rather two experimental runs $\endgroup$ – Angel Sep 19 '19 at 11:05
3
$\begingroup$

No. If the only difference between Tumour_batch1 and Tumour_batch2 was the library prep/sequencing run, then all this will tell you is if the particular sample of RNA taken from the tumour is different from the particular sample of RNA taken from normal. It almost certainly is.

Indeed if you did technical replicates of two RNA samples, both taken from normal, you would almost certainly find differences.

What you want to find is the differences in RNA level between tumour in general and normal in general. To do that you need to make an estimate of the mean and variance of RNA levels in the tumour and in the normal. To do that you need multiple samples from each.

If you need differences within a single patient, it would be more honest just to compute the fold changes without any implication of statistical certainty. You would then need to think about alternate schemes for measuring the confidence in inferences drawn from the data when you consider the actual biological questions.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.