advice for assessing concordance between primary capture and reciprocal capture results

[Former user]

Hello

I was hoping for a few insights into how best to compare the results from our primary capture Hi-C experiment, and a reciprocal validation capture. The reciprocal experiment contained ~4000 baits.

As it is not advisable to compare the sets of stringent interactions directly, I have tried using the sdef package with published data. I've tried to reproduce the validation results presented in Javierre et al., but I do not arrive at perfectly agreeing values.

So my questions are: To prepare the matrix for input into sdef, should I include only the interactions which occur in both experiments and create an ID such as 'bait1_otherEnd1' etc? And then should I assess interaction p values, weighted P values, or CHiCAGO scores?

thanks! JB

[Mikhail Spivakov]

We used all interactions that exceeded the score of zero (sic!) in at least one of the experiments (i.e., either primary or reciprocal) and exp(–Chicago_score) as input for baymod, setting p-value = TRUE. Note that we actually used a developmental version of this package, known as BGcom and available from here: http://www.bgx.org.uk/software.html. I'm realising only now that it's not exactly identical to the one in sdef, although the differences are minor - so thanks for bringing this up.