The invention discloses a restrictive two-stage
genome-wide association study (GWAS) method based on an SNPLDB mark, and aims at solving the problems that a traditional method cannot be used for estimating multiple
allele information, and is high in
false positive rate and low in efficiency of detection on
inbreeding crops. By combination of the SNPLDB mark constructed on the basis of a
haplotype block, correction of deviation of an
inbreeding population relation analysis model and a two-stage relation analysis strategy under a multi-
site model, the GWAS method suitable for conventional breeding of
inbreeding crops is built; the SNPLDB mark is applied to GWAS, so that a method is provided for multiple
allele estimation; candidate sites are screened on the basis of a single-
site model in the first stage, and are further screened on the basis of a progressive
regression analysis method in the multi-
site model in the second stage, so as to balance the problems of missing of
heritability and over-high
heritability estimation. Therefore, the interpretation ratio of a final
genetic model is controlled at trait
heritability. The positioning accuracy and efficiency are improved by a
feature vector and an appropriate significant level of a similarity
coefficient matrix estimated by the SNPLDB mark by GWAS.