Novel Machine Learning Approaches in Image-based Host-pathogen Interactions Analysis

The research of interactions between the pathogens and their hosts is key for understanding the biology of infection. Commencing on the level of individual molecules, these interactions define the behavior of infectious agents and the outcomes they elicit. Discovery of host-pathogen interactions (HPIs) conventionally involves a stepwise laborious research process. However novel computational approaches including machine learning and deep learning allow to significantly accelerate the discovery process, particularly for rich information sources like microscopy. One example of such approaches includes an algorithm we have devised to detect interactions between intracellular Toxoplasma gondii parasites and the host cell innate immune response molecules with high accuracy from micrographs obtained in high-content fashion. In another example, it was possible to detect intracellular and extracellular poxvirus virions in 3D superresolution micrographs without specific immunohistochemical labelling. This was possible through transfer-learning-enabled deep learning model inference from seemingly irrelevant fluorescence channels, allowing to distinguish minute changes in virus particle signal upon internalization. Finally, bringing temporal dimension as a source of information for deep learning algorithms allows predicting infection outcomes in a population of infected and uninfected host cells employing time-lapse microscopy data. Altogether, these examples suggest a great potential for HPI analysis using novel machine learning.