I am using SLiM to simulate slightly deleterious mutations in humans under a gamma distribution. From the literature I know that I can use a shape parameter of 0.23 and a mean parameter such that NeS=1000 (Eyre-Walker, Woolfit & Phelps, Genetics, 2006). However when Ne=100, this would leave a mean selection coefficient of 10, which is definitely not slightly deleterious. In fact this would prove strongly deleterious. Am I missing something here?