Corset p- values for threshold determination

[Harris Fisher]

Hello,

Is there a way to have Corset output a file of the p values from the likelihood tests in order to better determine appropriate thresholds or how well the data fit the Poisson model? Can one specify a zero-inflated poisson model for data that have a lot of zero counts?

Best,

- Harris 

[Nadia Davidson]

Hi Harris,

The likelihood test in corset is very simple as it's performed a large number of times and needs to be fast, so it's not possible to output the p-values or modify the model. However, the source code is open source, so if really would like to do this it's still possible by altering the code yourself and recompiling. 

When you say your data have a lot of zero counts, do you mean you have many samples with mostly zeros for a particular contig and then a few samples with non-zeros? Something like single cell data? Or do you mean that many contigs have zero counts even when summed around samples? This later scenario would be handled well in Corset already because those contigs will either be remove in early filtered or will be clustered on their own and won't undergo the Likelihood test. The early scenario I might need to think about a bit more.

Cheers,

Nadia.