Optimal estimation of transposition rates of insertion sequences for molecular epidemiology

Outbreaks of infectious disease can be confirmed by identifying clusters of DNA finger-prints among bacterial isolates from infected individuals. This procedure makes assumptions about the underlying properties of the genetic marker used for fingerprinting. In particular, it requires that each finge rprint changes sufficiently slowly within an individual that isolates from separate individuals infected by the same strain will exhibit similar or id entical fingerprints. We propose a model for the probability that an indivi dual's fingerprint will change over a given period of time. We use this mod el together with published data in order to estimate the fingerprint change rate for IS6110 in human tuberculosis, obtaining a value of 0.0139 changes per copy per year. Although we focus on insertion sequences (IS), our meth od applies to other fingerprinting techniques such as pulsed-field gel elec trophoresis (PFGE). We suggest sampling intervals that produce the least er ror in estimates of the fingerprint change rate, as well as sample sizes th at achieve specified levels of error in the estimate. Copyright (C) 2001 Jo hn Wiley & Sons, Ltd.