Determining sample size for co-expression analysis

Question

[daniel.cole.w]

Hi all,

I am looking into performing weighted co-expression like analysis on methylation data. Say ech sample has 10000 measurements, is there a guideline for how many samples I will need to construct meaningful co-expression networks of these measurements?

Any literature on the subject is appreciated.

Thank you, Dan

Answer 1

[Khader Shameer]

Like any quantitative biological data analytics, for WCGNA higher number of samples gives robust co-expression clusters.

WCGNA recommends minimum 15 samples. However, I would recommend you to do a power calculation and perform appropriate filtering of probes/genes to derive optimal number of modules.

[WouterDeCoster]

Is there any way to do a power calculation without a priori knowing/estimating the effect size?