Genetic analysis of francisellosis field outbreak in Atlantic cod (Gadus morhua L.) using an ordinal threshold model

Abstract

Ordinal threshold (OTH) models are commonly used for analysis of categorical data with logically ordered discrete response variables of a fixed number of classes, which follows a multinomial distribution. In this study, data from a natural field outbreak of francisellosis caused by the bacterium Francisella noatunensis in farmed Atlantic cod ( Gadus morhua L.) from the national Norwegian cod breeding program was analyzed. Mortality was recorded for a period of 81 days from the first day of the disease outbreak. Overall mortality at the end of the recording period was 24%. Surviving fish were dissected and infection statuses of the survivors were recorded. Based on this, disease resistance was scored in three distinct ordered categories; dead (1), alive but infected (2), and alive and uninfected (3). The data comprised of 2370 individuals from 176 families. The few fish that were alive and categorized as uninfected at the end of the test ( n = 82) were confirmed to be pathogen free by enzyme-linked immunosorbent assay (ELISA) and lack of growth of F. noatunensis from head kidney. The data were analyzed using an OTH with three different categories and predictive ability of this model was compared with a binary threshold (BTH) model (dead/alive) and a longitudinal survival score (SS) model for time-until-death. The estimated underlying heritability for liability to francisellosis was moderate to low for BTH (0.17 ± 0.08), OTH (0.14 ± 0.07) and SS (0.10 ± 0.05). Pearson correlation coefficients between mid-parent estimated breeding values (EBV's) for the different models were all close to unity. Predictive ability of the models, measured as square root of Pearson correlations, between mid-parent EBV's from two independent sub-sets of the data, indicated that the highest accuracy of family selection would be obtained with the OTH (0.73) followed by SS (0.72) and BTH (0.70). In conclusion, splitting survivors into subclasses, depending on infection status, did add additional information to the genetic evaluation. Hence, it is recommended to discriminate between infection statuses among survivors in genetic evaluation of francisellosis survival data. • Francisellosis disease data was analyzed using an ordinal threshold model. • We also compared disease data analysis with different statistical models. • We reported francisellosis disease field outbreak in adult Atlantic cod. • Ordinal threshold models gave highest accuracy for francisellosis data. • Useful results for Atlantic cod breeding program and other fish breeding programs