AlphaFold and related Artificial Intelligence (AI) methods now predict the structures of many protein complexes with near-experimental accuracy, and they are increasingly used to model how pathogen proteins interact with host proteins. However, for host-pathogen interactions, the success rate remains low. In this webinar, we will explain how AlphaFold-style methods predict protein complexes, and why they depend on evolutionary information that host and pathogen protein pairs largely do not provide. We will then go through what recent large-scale studies have achieved, where the predictions worked and where they failed, and how the reported success rates should be read. Finally, we will describe strategies that can improve the results, including extensive sampling, modified sequence alignments, and the use of experimental data such as crosslinking mass spectrometry, illustrated with our work on influenza A virus.This event is part of the webinar series “Integrating structural biology and bioinformatics to study infection”. You can follow the link for more information about the series and its webinars.