Vary the representation of patrilines (fathers), possibly with difference between workers and drones

[gregorgorjanc]

Cryptic “royal” subfamilies in honey bee (Apis mellifera) colonies https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0199124

... Through patriline analysis of colonies along with large numbers of emergency queens reared by each we affirm the purported “royal” patriline theory that, instead of competing nepotistically, workers exhibit bias towards selecting individuals from particular “royal” subfamilies during emergency queen rearing events.

Further, we show that these “royal” patrilines are cryptic in honey bee colonies; occurring in such low frequency in the overall colony population that they are frequently undetected in traditional tests of queen mating number and colony composition.

The identification of these cryptic “royal” subfamilies reveals that honey bee queens, already considered “hyperpolyandrous,” are mating with even more males than has been previously recognized.

[gregorgorjanc]

Of note, in beeCross() (the haply-diploid version! - see https://github.com/HighlanderLab/SIMplyBee/pull/142) we call randCross2(..., balance = FALSE). We could use balance = TRUE if needed? But, the main point with the above paper is that distribution of pari-lines between workers and virgin queens is different, so we could/should be looking at "balancing" differently for workers and virgin queens. If we do this.

[gregorgorjanc]

We were discussing today that at the moment we sample fathers at random when creating new workers or virgin queens. This will add some noise in how fathers are used, but we expect fairly uniform usage. Say for 100 workers from 3 fathers A, B and C we might see A as father 39x, B 30x and C 31x.

We could add a way for a user to pass a vector of probabilities to enable much different use of fathers, say 10, 60, 30 for the above example. We could sample proportions for such a vector from the Dirichlet distribution. The challenge here is in getting good/reasonable default parameters from publications. Maybe the above paper is already useful!

We could add the Dirichlet distribution function to SimParamBee - like we do with n and p functions.

We could show how such probabilities would be a function of drones phenotype for trait sperm production;)

Then we can make this vector into a matrix with two rows and one is for workers and another for virgin queens to vary use of fathers for each caste.

When creating virgin queens for artificial queen rearing we will likely have to use "worker" probabilities and for swarm and supersedure "virgin queen" probabilities.