M1 family size?

[S. Cutler]

  Dear Arabidopsis Community,  

	My name is Sean Cutler, I am a graduate student of Peter McCourt's  
at the University of Toronto. I'm currently screening for mutations 
which affect seed development in Arabidopsis. I have a question which 
I've been  thinking about since I started screening for mutants, and I 
thought I'd put it out to the community to see what people think. Before 
I ask however, I thought I should pretext the question with a statement 
that I understand that there is no 'right' answer to the question; I ask 
because I'm curious as to what other people think and do. 
	So here it is. It's quite simple. How many M1 do you think should be  
used in a screen to identify new genes?I know the safe answer.
	If you look at Koorneef's data from 1982 (Mutation research) he  
showed that mutation frequency per locus (using 10 mM EMS, 24 hrs.)   
varies in Arabidopsis from  about 1 in 20,000 for gl type mutations  to 
as high as 1 in 2,400 for several ga genes. The average rate was around 
1 in  5000.   
	Based on a Poisson distribution of mutation events, you should 
screen an M1 family large enough to theoretically contain at least 5 
mutants  in a gene to be 99% certain that you will observe that mutant 
at least once. If you assume that 1 in 20,000 is a safe number for the 
lowest EMS  mutation rate, then the 'safe' way to screen would be to 
screen M2 seed  from an M1 family of at least 100,000 M1 plants. With a 
genetically effective cell number of two , this drops to about 50,000.  
(the GECN  allegedly varies, so perhaps a GECN of 1 should be assumed 
for safety?)  
	To screen M2 seed from this many M1 is a large task. How 
do people  generally deal with this? Should a screen which is going to 
be published  be done to saturating limits? I know the answers to these 
questions can  be very personal, but I'd like to know how people feel 
about this.  

	I'll be happy to summarize the responses to my question.  

Thanks.  

Sean Cutler:   cutler at gpu.utcc.utoronto.ca