Large scale recording of phenotypes and DNA information under commercial pond environment is critical for implementing efficient breeding programs in aquaculture species. DNA information can help to overcome the limitations and challenges of pedigree recording and individual animal tagging. In addition, recent development in sensor technologies and imaging combined with pattern recognition are providing the avenues for developing the high-throughput recording systems in aquaculture. We generated a large resource population of black tiger shrimp (Penaeus monodon) using high-throughput phenotyping and genotyping for genetic studies and gene discovery, and implementing genomic selection, as part of the ARC Industrial Transformation Research Hub for Advanced Prawn Breeding, an Australian organisation consisting of a partnership among universities, CSIRO, and the private sector. This resource population consisted of 400 families of varying sizes across three generations.
We developed a custom-made phenomics system for recording images of shrimps in batches. This system was used to record digital images on a large number of animals (>30,000) at harvest under commercial pond environments. We also recorded manual weights and components traits of a proportion of these samples for constructing a training dataset for deep learning models. We will describe examples of high-throughput recording and accuracy of predicting of phenotypes from RBG images using deep learning models. We will discuss the results of analysis of genetic architecture of body size traits, accuracy of genomic prediction within and across generations, and different factors affecting the accuracy of genomic selection in black tiger shrimp.