TP, FP, TN and FN in error correction reads

Question

[midox]

Hello,

I want to know how to calculate the TP, FP, TN and FN in the case of error correction?

Thanks

Answer 1

[Devon Ryan]

The same way you would calculate them in any case, with a dataset that has known correct results that's then run through the procedure.

[midox]

I don't have these results. I have just the Reference genome. Do I use the reference genome?

thanks

[Devon Ryan]

Yes, you would have to use a reference genome to generate fake data with known properties what would then be error corrected. Ideally you just sequence the reads, but that assumes that the reference is correct (maybe it's a good assumption, maybe not).

[midox]

Or the best metric is the mapping of corrected reads to the reference genome. Am I right?

thanks

[Devon Ryan]

No, then you're adding another layer of processing and confounding things by how well the aligner is able to deal with the reads and the genome.

[midox]

I do not have a choice. Otherwise how I validate my corrected reads??

[Devon Ryan]

You don't

[midox]

I have the reference genome. So I can map my corrected reads to the reference genome and I see the results of mapping.