miRNA-Target coexpression and network

Question

[Prasad]

Hello, i have miRNA data for 4 samples (plant whose genome is not available) and have corresponding predicted targets. I have the following queries.

1. What is the way to the co-expression analysis between miRNA and its targets.
2. How to generate the miRNA and target network based on the expression data

Any suggestions. Thanks

[Michael]

I think this is really going to be a bit difficult, and you will need a bit more than the 4 miRNA samples. I think first step is to find out if you have sufficient data at all and if it is worth doing the analysis. So you would need: assembled transcripts, seems like you have them because have predicted targets. You definitely also need RNA-seq data, you also mention expression data, would optimally be replicated.

The simplest way of determining co-expression between miRNA and target is based on correlation (target expression should be negatively correlated with miRNA expression in theory), as is the construction of a co-expression networks. The downsides are in my opinion, the target prediction for miRNA is really noisy with many false positives, and it is not clear if RNA-seq can capture well the post-transcriptional regulation by miRNA. In addition, I am assuming that the 4 samples also were used to predict miRNA's de-novo, now these should be re-used for differential expression analysis. The whole setting bears very little foundation in validated structural features (de novo genes/transcripts, de novo miRNAs, de novo targets for these, gene expression analysis using the same data used to build transcripts and miRNA). I think we could pull up a few papers arguing for either position.