Although quantitative trait loci (QTL) underpin many desirable agronomic traits, their incorporation into crop plants through marker-assisted selection is limited by the low predictive value of markers on phenotypic performance. Here we used candidate defense response (DR) genes to dissect quantitative resistance in rice using recombinant inbred (RI) and advanced backcross (BC) populations derived from a blast-resistant cultivar, Sanhuangzhan 2 (SHZ-2). Based on DNA profiles of DR genes, RI lines were clustered into two groups corresponding to level of resistance. Five DR genes, encoding putative oxalate oxidase, dehydrin, PR-1, chitinase, and 14-3-3 protein, accounted for 30.0, 23.0, 15.8, 6.7, and 5.5% of diseased leaf area (DLA) variation, respectively. Together, they accounted for 60.3% of the DLA variation and co-localized with resistance QTL identified by interval mapping. Average phenotypic contributions of oxalate oxidase, dehydrin, PR-1, chitinase, and 14-3-3 protein in BC lines were 26.1, 19.0, 18.0, 11.5, and 10.6%, respectively, across environments. Advanced BC lines with four to five effective DR genes showed enhanced resistance under high disease pressure in field tests. Our results demonstrate that the use of natural variation in a few candidate genes can solve a long-standing problem in rice production and has the potential to address other problems involving complex traits.