sharing information across sites

[jbloom]

The idea here is that if we have multiple sera, we might want to constrain them to have similar epitopes even if activities are different and exact beta values change a bit.

Right now we think this idea should be kept in mind, but we should wait to fit a model on real serum data before starting too much on this.