Next level virus isolation in cell culture using artificial intelligence

Background:

Virus isolation from mammalian cell culture is still the only method of propagating novel viruses for further characterisation. For this purpose, treated cell cultures are observed microscopically on a daily basis over a period of at least 7 days as viruses often cause cytopathic effects (CPEs) in cell culture. If CPEs are observed under the microscope, the cells and the supernatant are collected and further analysed by specific PCRs, NGS or electron microscopy. However, a CPE is not always clearly visible to the human observer. Additionally, certain viruses, e.g. Ebola virus, hantaviruses, etc., are known to cause no visible CPE in infected cell cultures whatsoever, despite those viruses multiply in the cells to high titres. As a result, studies to isolate as yet unknown viruses often involve hundreds of samples, and so far, only the few (~5 %) that show CPEs visible to the human observer are analysed further. Since infection of cells always leads to changes within the cells though, it may be possible to detect these by computational methods using AI, even if they are not be clearly visible to the human eye.

Aim/s:

Within the project an algorithm will be trained to distinguish infected cell cultures from non-infected cell cultures. Based on this trained algorithm then viruses causing only limited or no visible CPE in cell culture will be included in the project. The overall aim is, to generate a tool which could be widely used to identify infected cell cultures independent of the detection of visible CPE. The tool would have the potential to increase the virus isolation rate, allow for the detection of yet unknown viruses and to speed up the whole isolation process. The outcome of this study can be applied to several similar research questions in relation to the identification of viruses using imaging techniques.

AI methods:

To identify infections in cell cultures caused by different viruses, deep learning models such as convolutional neural networks and vision transformers have to be implemented and trained.

Keywords:

Artificial Intelligence, Computer Vision, Viral Infections, Deep Learning, Public Health

to the top