World Aquaculture Singapore 2026

June 2-5, 2026

Singapore

SEX PREDICTION USING GENOMIC INFORMATION AND MACHINE LEARNING IN SOCKEYE SALMON Oncorhynchus nerka

Jane E. Symonds*, Roy Costilla, Rudiger Brauning, Shannon M. Clarke, Tracey van Stijn, Eva Wilson, Jon Bailey, Jaco Swart, Glenn Tribe, Megan R. Scholtens, and Ken G. Dodds.

 

Cawthron Institute

98 Halifax Street East

Nelson 7010

New Zealand

jane.symonds@cawthron.org.nz



Current evidence shows that both genetic and environmental factors affect the expression of phenotypic sex in salmonids, including Sockeye salmon (Oncorhynchus nerka). This species was successfully introduced to New Zealand over 100 years ago and wild populations remain in the South Island. However, questions remain about whether sex can be reliably predicted and the degree of discrepancies between genetic and phenotypic sex in these new environments. Here, we used genomic data from 373 samples to predict sex in New Zealand Sockeye salmon from three populations across two cohorts (2024 and 2025). First, we identified “sex markers” using a Fst test for SNPs from genotyping-by-sequencing (GBS) data from 2024, e.g. SNPs with diverging allele frequencies between males and females. Sex prediction was then made in the 2025 cohort using the sex markers with two methods: (1) extreme gradient boosting (XGBoost), a machine learning (ML) model, and (2) a regression model analogous to breed prediction.

We found that the sex markers were in LG9a and LG9b, at previously associated loci with sex determination (Figure 1). Prediction rates for females were close to 100%. Prediction rates for males were a bit lower ranging from 77% to 93%. Prediction rates using ML were very similar (Table 1). Our results are comparable to genetic tests for sex other salmonid species.

However, care is still needed when extrapolating results across other cohorts as sex prediction could be less accurate and this could delay the development of all-female broodstock for breeding programs. Overall, we demonstrate the feasibility of using genetic testing for sex using SNPs derived from GBS data in New Zealand Sockeye salmon.