By Rodrigo Godinho and Ina Ranberg
Topigs Norsvin routinely applies artificial intelligence in semen analyses to assist in the production of semen doses for artificial insemination. Sperm traits are assessed using CASA, Computer-Assisted Sperm Analysis systems, which is a standardized high-throughput phenotyping method that uses machine learning algorithms to assess the concentration of sperm cells in ejaculates. It also describes motility parameters, including progressiveness and morphology of individual sperm cells, yielding indexes of morphologically normal and abnormal sperm cells due to head or tail defects of the sperm cell and cytoplasmic droplets.
- In the “Big Data Male Reproduction project”, large datasets of over 1.5 million ejaculates from over 25,000 boars generated in multiple CASA systems in Canada, Germany, the Netherlands, and Norway are combined and sources of variation in semen quality are described. This information is connected to over 300,000 litter results at the ejaculate level, enabling the assessment of the influence of the boar on field fertility. Information on the boars’ early life conditions, such as puberty, libido and weight development, are also connected and evaluated.
- In the “Super Semen project”, a deeper dive is taken into CASA. Information of each of around 800 single sperm cells per ejaculate is recorded and stored directly in the data cloud. The effect of those single-cell data on field fertility is determined, in combination with information on seminal plasma components analyzed using spectroscopy. Special attention is given to the further evaluation of different methodologies, including cluster analysis and machine learning methods.
The general aim of these projects is using big data to optimize the reproduction capacity of males to improve field fertility.
The breeding program benefits because CASA automates a laborious and important task and enables opportunities for improvement in the artificial insemination business. The deliverables of these projects are improved quality of semen for all applications, increased accuracy of fertility predictions, improved standards of dose production, increased longevity of boars with the highest genetic potential in the studs, and increased capacity and training in machine learning and computational models.
These projects contribute to the pork production chain because they will enable more efficient dissemination of the breeding company’s germplasm. The more efficient use of boars will decrease the number of individuals needed and decrease the concentration of semen doses, resulting in increased sustainability of the entire chain.
