Experimental evaluation of several cycles of marker-assisted selection in maize

A program was initiated in 1994 to compare the efficiency of marker-assisted selection (MAS) and conventional phenotypic selection. A population of 300 F-3:4 families was generated from the cross between two maize inbred lines F2 and F252 and selected on an index combining grain yield and grain moisture at harvest. This population was characterised for 93 RFLP markers and evaluated as testcrosses in a large range of environments. Three methods of selection were applied (i) two cycles of conventional phenotypic selection; (ii) two cycles of MAS based on an index combining phenotypic values and QTL genetic values and (iii) one cycle of combined MAS followed by two cycles of selection based only on the QTL effects estimated in the first generation. The different populations were characterised for RFLP markers. The evolution of allele frequencies showed that selection on only-markers was very efficient for fixing QTL alleles found favourable in the initial population. This evolution was quite different from that observed for phenotypic selection or combined MAS. Genetic gain was evaluated and found significant for each method of selection. Nevertheless, the difference between phenotypic selection and combined MAS was not significant. The two additional cycles of MAS on only-markers did not improve significantly the genetic value of the population. Moreover, the genetic variance of this population remained high, despite most of the QTL initially detected were almost fixed. The results suggest that the QTL effects estimated in the initial population were not stable due to epistasis and/or QTL by environment interactions.