Comparison of kinship‐identification methods for robust stock assessment using close‐kin mark–recapture data for Pacific bluefin tuna

Abstract

Several attempts have been made to understand the population dynamics of fishery resources, such as tuna species using an integrated analysis model with multiple data sources. However, estimating absolute abundance levels in practical stock assessments remains a challenge. Close‐kin mark–recapture (CKMR) methods provide information about the number of adults in a population using close‐kinship pairs identified by genetic markers and statistical methods. In this study, we compared three methods for kinship identification using different algorithms in samples of wild Pacific bluefin tuna genotyped across 5029 genome‐wide single nucleotide polymorphisms in 4108 samples. The flexible relationship analyzer by random forest method we developed employs pairwise identity‐by‐descent values as inputs for random forest classification. The other two methods were CKMRsim and COLONY, which have been published and applied in several studies. These three methods were applied to the actual genotyping data with moderate missing genotypes, in addition to the pseudo‐generated genotyping data for the simulation test. The simulation test mimicked genotyping data with physical linkages as well as genetic characteristics similar to those of actual samples. The three methods resulted in different numbers of inferred kinship pairs for both generated and actual data. Particularly for the half‐sibling pairs, a considerable number of false‐positives and false‐negatives existed in the identification results. The differences in kinship identification results were interpreted based on a simulation test. This study may enhance the understanding of how each software performs when applied to single‐nucleotide polymorphism data with moderate missing genotypes, as demonstrated in this study.