What is the minimal sample size required to run MIIC?

[guodudou2]

Hello,

I am interested in using this package, but wondering what is the minimal sample size that is required for RNAseq data?

Thanks!

[franck-simon]

Hello guodudou2

Thanks for your interest in MIIC method.

It is difficult to give a minimal number of samples as it depends on the quality of your data and the strength of the effects.
To try to give you an order of magnitude, for 10X scRNAseq, that contains a lot of dropouts, a number of samples of some thousand(s) should be fine and with Smart-seq3, without dropouts, a number of samples of some hundred(s) can be enough.
Running MIIC with very few samples is almost always possible, just have in mind when interpreting the results, that the less samples you have, the less robust will be the findings.

Best regards

[guodudou2]

@franck-simon Thanks for the information. I would like to apply network analysis to bulk RNAseq data, ranging from a dozen to a hundred samples per group, do you think it is worth trying MIIC? If not, what tools do you recommend?

[louise-rb-dupuis]

Hello guodudou2,

Indeed a dozen of samples by group can be a bit short to reconstruct a MIIC network in general (meaning you have to be careful interpreting the network), but bulk RNAseq are what we see as "quality" samples, they have a lot of information that we can try to recover ! So I would say it's worth a go :) By reconstructing a MIIC network you can see what dynamics are at play in your system. To do so, you will first have to select genes of interest as you can't reconstruct a network on your 25000-ish genes. I would advise to select under 200, it can be from biological knowledge or unsupervised methods as you like. You could add your group variable as a metadata (and any other metadata of choice really) to see how it influences those dynamics.

Do not hesitate to reach if you have any more questions !
Best regards