Is is fair (or correct) to separate subjects in high responders or low responders based on expression value of a gene in an seqRNA experiment? In other words can we use per sample gene expression for anything? I thought we are looking at the data as groups. Thank you.
Not sure what you mean, but there are various disease sub-types where the heightened expression of a single gene is predictive of the sub-type, e.g., EGFR expression and 70% of lung cancers, ERBB2 / Her2 expression and Her2-positive breast cancer, etc.
I've been asked to use expression or fold change for gene x in patient A to compare with the patient B for the same gene. Both patients are part of a big study. I said that is possible but is not correct.
So, just comparing 2 numbers? - all that you can do is simple arithmetic:
I know the math. I don't believe it is correct to look at one patient Fold Change in a group. That was my original question/ doubt. It is meaningless.
You can't say they are significantly different. Either or both of those samples could be outliers. That does not mean the values are meaningless. Outliers are a problem with any classification problem. Does it mean we can't classify individual samples?
Thank you Igor for agreeing that you cannot say that they are significantly different. Each of them is part of the group, its value makes up for the group value, if they are outlier we can eliminate them. I can see the number difference in expression for one gene but I cannot infer from that that this sample (patient) responds better than the other one. At least this is my understanding, I might be wrong, this is why I asked. I do not think we can classify samples based on expression gene levels or fold change. We can use other metrics (like library size, number of exons, etc) but not this one.
You can certainly classify samples based on gene expression. Classic example is PAM50, which has been used clinically for over a decade: https://ascopubs.org/doi/10.1200/JCO.2008.18.1370