Reading this [@Wang:2010fh], in particular Supp. report 4. Has to do with bootstrapping population estimates using the different enzymes used to find integration sites.
The review [here @Chao:2001el] indicates sample coverage approaches may be most appropriate for models where the capture probabilities can vary both over time and between species (which is certainly the case for estimating populations from longitudinal T cell sequencing data). I’ll need to read more (esp. [@Tsay:2001vc] and [@Chao:1998ks]).
Also found the package Rcapture which appears to fit a Poisson regression to incident counts. I’ve gotten it to work (kind of) but R crashed trying to plot the model so I’ll need to come back to this. If memory serves, it was substantially more conservative in its estimates than the Chao2 estimate.