Evaluation of methods for the estimation of mutation rates in cultured mammalian cell populations

A systematic comparison of 5 different statistical methods for the estimation of mutation rate (mu) in cultured Chinese hamster V79 cells is presented. Fluctuation tests were performed with several large batches of parallel cell cultures each allowed to grow for a different length of time in order to reach different population size (Nt). Based on Lea and Coulson's theoretical distribution, a comparison has been made between the experimental data and the expected distribution of the number of ouabain-resistant mutants per culture in these hamster cell populations. The sum of squared deviation between the observed and expected values, or SSD, was used as a means of the adequacy of the estimation method; the method which gives the smallest SSD is regarded as the best one for the estimation of mu. Our results show that when Nt is small, the occurrence of mutation is infrequent, and SSDs from different methods are similar. However, when Nt is large, there is a great discrepancy of the SSD values, suggesting a preference of using the maximum likelihood method, the Po method, the median method, the upper quartile method and the mean method, in that order, for the estimation of mu. The order of preference is correlated with estimation efficiencies. Depending on the size of Nt and the method used, the estimated mu may vary up to more than 3-fold. At a large Nt, the mu obtained from the maximum likelihood method is very precise. This suggests the importance of choosing an appropriate Nt as well as method for the estimation of mu.