1Experimental evolution is becoming a popular approach to study genomic selection responses 2 of evolving populations. Computer simulation studies suggested that the accuracy of the sig-3 nature increases with the duration of the experiment. Since some assumptions of the com-4 puter simulations may be violated, it is important to scrutinize the influence of the experimen-5 tal duration with real data. Here, we use a highly replicated Evolve and Resequence study in 6 Drosophila simulans to compare the selection targets inferred at different time points. At each 7 time point approximately the same number of SNPs deviated from neutral expectations, but 8 only 10 % of the selected haplotype blocks identified from the full data set could be detected in 9 the first 20 generations. Those haplotype blocks that emerged already after 20 generations dif-10 fer from the others by being strongly selected at the beginning of the experiment and displaying 11 a more parallel selection response. Consistent with previous computer simulations, our results 12 confirm that only Evolve and Resequence experiments with a sufficient number of generations 13 can characterize complex adaptive architectures. 14 15 17 18 Deciphering the adaptive architecture is a long-term goal in evolutionary biology. In contrast 19 to natural populations, experimental evolution (EE) provides the possibility to replicate exper-20 iments under controlled, identical conditions and to study how evolution shapes populations 21 in real time (Kawecki et al. (2012); Schlötterer et al. (2015)). The combination of EE with next-22 23 (2015); Long et al. ( 2015)) -has become a popular approach to study the genomic response to se-24 lection and to identify adaptive loci. E&R has been applied to various selection regimes, such as 25 virus infection (Martins et al. (2014)), host-pathogen co-adaptation (Papkou et al. (2019)), ther-26 mal adaptation (Orozco-Terwengel et al. (2012); Barghi et al. (2019)), or body weight (Johansson 27 et al. (2010)). A wide range of experimental designs have been used, which vary in census popu-28 lation size, replication level, history of the ancestral populations, selection regime, and number and Hughes (2018); Turner and Miller (2012); Rêgo et al. (2019)), over a few dozen (Orozco-33 Terwengel et al. (2012); Johansson et al. (2010)), up to hundreds of generations (Burke et al. 34 (2010)). Computer simulations showed that the number of generations has a strong influence 35 on the power of E&R studies, and increasing the number of generations typically improved the 36 results (Baldwin-Brown et al. (2014); Kofler and Schlötterer (2014); Vlachos and Kofler (2019)). 37Since simulations make simplifying assumptions, it is important to scrutinize these conclu-38 sions with empirical data. Until recently no suitable data-sets were available, which included 39 multiple time points and replicates. We use an E&R experiment (Barghi et al. (2019)), which 40 reports allele frequency changes in 10 replicates over 60 generations in 10 generation interval...

Evolve and Resequence (E&R) studies investigate the genomic selection response of populations in an Experimental Evolution (EE) setup. Despite the popularity of E&R, empirical studies typically suffer from an excess of candidate loci due to linkage, and single gene or SNP resolution is the exception rather than the rule. Recently, a secondary E&R design – where unevolved founder genotypes are added to a primary E&R study - has been suggested as promising experimental procedure to confirm putative selection targets. Furthermore, secondary E&R provides the opportunity to increase mapping resolution by allowing for additional recombination events, which separate the selection target from neutral hitchhikers. Here, we use computer simulations to assess the effect of crossing scheme, population size, experimental duration, and number of replicates on the power, and resolution of secondary E&R. We find that the crossing scheme, population size, and the experimental duration are crucial factors for the power and resolution of secondary E&R: a simple crossing scheme with few founder lines consistently outcompetes crossing schemes where evolved populations from a primary E&R experiment are mixed with a complex ancestral founder population. Regardless of the experimental design tested, a population size of at least 4,800 individuals, which is roughly 5 times larger than population sizes in typical E&R studies, is required to achieve a power of at least 75%. Our study provides an important step towards improved experimental designs aiming to characterize causative SNPs in EE studies.