PROSPECTS FOR INCREASING GENETIC GAIN IN PLANTATION FORESTS THROUGH MARKER-AIDED SELECTION - RESULTS OF SIMULATION STUDIES

New Zealand Forest Research Institute, Private Bag 3020, Rotorua, New Zealand

Use of cosegregation analyses to identify marker-trait associations for subsequent marker-aided selection (MAS) was expected to be straightforward, but MAS has not yet been demonstrated to be successful for quantitatively inherited traits of forest trees. Tests of marker-trait associations on independent verification populations have repeatedly indicated that associations were not nearly as large as first indicated, and suggest that many may not exist at all. Statistical approaches to identifying marker-trait associations appear to work well when gene effects are large, but can lead to erroneous conclusions when effects are small, as appears to be the case with forest trees. In order to be confident enough to use markers identified solely through cosegregation analyses for MAS in forest trees, simulation is being used to obtain a better understanding of the pitfalls inherent in QTL detection. The simulations highlight the limitations of the statistical methods used to date, but suggest that experimental designs for cost effective MAS can be developed. Specifically, simulation studies with small effect genes suggest: 1. The relative power of alternate methods of cosegregation analyses indicates relative gain from MAS, but does not provide enough information to evaluate the utility of the methods. The number of markers chosen for MAS must also be taken into account because of the expense of obtaining marker information and generating progeny for selection. 2. Using markers where test statistics from single marker analyses are at a peak appears to be a cost effective approach to choosing markers for MAS. Although the gain (power) is slightly reduced when compared with interval mapping and composite interval mapping, it can be achieved with a substantially smaller number of markers, and thus a much reduced cost. 3. Use of experiment-wise thresholds (in contrast to marker-wise thresholds) and use of verification populations do not always have the desired effect of discarding only false marker-trait associations. 4. Selective genotyping of large populations appears to be a promising approach for forest trees.