Durability of blast resistance is a major concern in the upland ecosystem. Partial resistance may be more durable than complete resistance because it appears to be largely race-nonspecific. It cannot be assessed, however, unless one has a virulent strain effective against major genes. An isolate compatible with the highest possible number of parents may be used to overcome major genes in a population, which is then purged of the remaining major gene(s) conditioning resistance to such an isolate. Only the major genes efficient against the chosen isolate are discarded and the other major genes retained. Once such a population is built, lines with a susceptible infection type but having a reduced relative infection efficiency after inoculation with the compatible isolate can easily be selected. The effective major genes that were purged from the population can be introduced at the end of the program in order to have the best partial resistance and the best major genes. GPIR-22 is a poly-cytoplasmic gene pool we constituted following this strategy.