Abstract: 【Objective】 The current standards for plant variety identification only examine a small number of markers on the genome, which may lead to sampling errors, therefore identification conclusions are often questionable. The goal of this study is to estimate the sampling errors in plant variety identification procedure and evaluate the reliability of the conclusions, and eventually provide the scientific basis for the applications of molecular identification standards. 【Method】 Based on the number of observed differential loci between two varieties, a conditional probability model was established based on the Bayes’ theorem to estimate the true number of the different loci. Given that the observed number of differential loci between two varieties follows the binomial distribution, and the prior distribution of true number is an approximate uniform distribution, the conditional probability model was finally computed. Based on the confidence levels provided by the probabilistic model, the relationship between plant varieties is divided into the red, green or yellow zones, corresponding to the same or similar, different and undetermined varieties, respectively. To validate this probabilistic model, the genotyping data of 3 205 SSR molecular markers for each of the 8 rice varieties were used. For each pair of varieties, 10 000 sets of molecular markers were simulated, and each set is composed of 48 random SSR markers. For each simulation, the relationship between the varieties was estimated based on the probability computed by the model. And then the estimated relationship was compared with the real one to evaluate the accuracy of the probabilistic model. Finally, the probabilistic model was applied to provide probabilistic support for the conclusion of a recent watermelon variety infringement case. 【Result】The validation results showed that each pair of varieties was determined as different varieties in 4 295-10 000 simulations at a confidence level of 95%. Compared with the true relationship between varieties, the probabilistic model had an accuracy of 100% in determination of rice variety relationships. Finally, the court decision about the infringement dispute of watermelon varieties also was supported by the probabilistic model at a confidence level of 95%, indicating that the losing party's doubt on the limited number of the sampling loci is not sufficient. 【Conclusion】In this study, a probabilistic model was built to evaluation the reliability of the conclusion of the variety relationships, which provides confidence levels for the molecular identification conclusion of the relationship among varieties, and thus improves the accuracy, and finally avoids the controversies caused by the insufficient number of testing markers.

Key words: variety identification, molecular marker, sampling error