Genetic analysis of individual feed intake and efficiency in Atlantic salmon smolts using X-ray imaging

Sammendrag

Improving feed efficiency (FE) is poised to become a primary focus of future aquaculture breeding programs, with ongoing research aiming to incorporate this trait to increase fish production, reduce feed costs and minimize environmental impact. Selection for improving FE requires a detailed understanding of its genetic parameters and association with growth traits under commercial conditions. However, scientific studies on FE at the individual level are missing in Atlantic salmon, likely due to the absence of accurate and efficient methods for measuring individual feed intake (IFI). Here we have used X-ray imaging on 700 salmon smolts, which have consumed feed containing radio-opaque beads, to phenotype the IFI at multiple time points. The study aims to I) train and validate a deep learning image based “YOLO-XBeads” model for accurate estimation of IFI from X-ray images, II) estimate the genetic parameters of IFI and different FE metrics such as feed conversion ratio (FCR), phenotypic residual feed intake (pRFI), genetic residual feed intake (gRFI), and growth traits such as Body weight (BW), average daily gain (ADG) and whole-body fat (WBF), and III) quantify the phenotypic and genetic relationships among these traits in Atlantic salmon. The YOLO-XBeads model performed well on X-ray images, with a high correlation to the true value (R2 = 0.99) and mean absolute percentage error of 2.60 – 5.94 %. The time efficiency (two orders of magnitude faster than available software) of our model significantly reduces the analysis time, labour and cost compared to conventional image software and manual human counting. We found significant heritability estimates for IFI (0.20 ± 0.05 – 0.50 ± 0.07), ADG (0.44 ± 0.06 – 0.55 ± 0.06), BW (0.50 ± 0.06 – 0.55 ± 0.06), pRFI (0.11 ± 0.05 – 0.16 ± 0.07), gRFI (0.06 ± 0.03 – 0.18 ± 0.06), FCR (0.09 ± 0.04 – 0.23 ± 0.06) and WBF (0.61 ± 0.06), indicating potential for genetic improvement. Additionally, the strong genetic correlations (rg = 0.71 ± 0.07 – 0.98 ± 0.01) between different point estimates of the primary traits (IFI, ADG and BW) over time demonstrated the stability of these traits over the studied growth period. Averaging the FE and related traits over the whole growth period (50 – 300 g) of salmon showed variable rg between traits. The highest rg values (0.86 ± 0.05 – 0.99 ± 0.01) were observed among different FE metrics, indicating similar genetic regulation. Following this, a strong rg (0.95 ± 0.02) was found between IFI and ADG, indicating that ADG can explain a significant proportion of the genetic variation in IFI. However, we found low but significant genetic variation for pRFI and gRFI, which shows heritable variation remains after accounting for the relationships of BW and ADG with the IFI. In addition, WBF showed moderate rg (0.52 ± 0.09 – 0.64 ± 0.07) with IFI, BW and ADG traits. Conversely, FE metrics showed weak rg with ADG (0.00 ± 0.21 – 0.38 ± 0.17) as well as with WBF (−0.03 ± 0.18 – - 0.21 ± 0.19), indicating little or no indirect genetic improvement in FE through selection for growth or WBF in freshwater smolts. Overall, the FE and related traits were found to be measurable and heritable at the individual level in Atlantic salmon smolts, signifying the potential for improving economic gain and minimizing environmental impact. However, further studies are needed at the sea phase.