When we do a genetic screen, how can we calculate the haploid genome coverage, and how can we know if we saturate the genome?
Calculating haploid genome coverage is fairly straightforward, and only really depends on what generation you are looking in and whether you are looking for recessive, dominant, or maternal-effect phenotypes.
The most common case is the F2 nonclonal screen, in which you mutagenize, let self for two generations, and look in the F2 for homozygous animals showing a recessive phenotype. Each F1 contains two independent mutagenized genomes, from the mutagenized sperm and the indepedently mutagenized oocyte; if we call the two mutagenized genotypes at one arbitrary point in the genome “A” and “B”, then the resulting self-progeny F2 is 1/4 likely to be A/A at this locus, 1/2 likely to be A/B or B/A, and 1/4 likely to be B/B; thus, each F2 has a 50% chance to be homozygous for one of the two available mutagenized genomes at any given locus, and for recessive phenotypes you can take the number of F2s and divide by two to determine the number of homozygous mutagenized genomes screened. If you were looking for dominants, each F1 (if screened) would be worth two genomes, and each F2 would be worth 1.5 genomes. I’m sure you can work out similar formulae for other screening methods (non-complementation, maternal-effect, clonal, etcetera).
To maximize screening efficiency (and to make the math simpler by letting you basically disregard resampling), have to make sure you don’t spend a lot of time re-screening the same genomes: if you have one F1, it can only give you two genomes no matter how many of its F2s you screen. The usual practice is to have independent pools of mutagenized P0s, and screen a number of F2s from each pool that is several-fold smaller than the number of F1s that pool produces; I usually aim for not more than about 1/5 to 1/10. A good example of calculating the number of genomes screened - one that includes the sampling issue - can be seen in Ron Ellis’s engulfment paper.
Assessing saturation is harder, and can’t really be done by simple reference to the number of genomes screened. The usual estimate is that about 1 in 2000 EMS-mutagenized genomes will contain a loss-of-function mutation in the average gene (a good example of the data underlying this estimate can be seen in one of Iva Greenwald’s early Genetics papers). But of course your genes may not all be average, and simple loss-of-function mutations may not work for some genes you care about. Also, you may have to take in account issues of penetrance, how well you recognize the phenotype while screening, partial maternal rescue, etcetera. The usual approach is to do a screen, determine the number of genes and the number of alleles of each of them; then, if you have multiple alleles of each of them, consider yourself to have likely saturated or come close to it. Clearly, this is not a perfect approach. Using the same data, you can assess how many genes you’ve likely missed using the Poisson distribution, but remember that the Poisson makes the flawed assumption that all genes are equally mutable for your phenotype.