Deploying protein machine learning models on the web

Tuesday March 28th, 4-5 pm EST | Simon Dürr — PhD Candidate, EPFL (Switzerland)

In recent years, the field of machine learning applied to proteins has exploded with biochemists and the machine learning community contributing many exciting advances in terms of model architectures and used datasets. Very successful models such as AlphaFold2 have benefited from being available as easy to use webapps (ColabFold) on services like Google Colab with integrated visualization of the output and configurable input options. While extremely successful it is currently not standard to share such interactive demos or even code at the preprint stage hindering progress in the field.   In this talk, I want to give an overview of the landscape of tools to publish such models including a live demo of how to convert a model from a python script to an interactive webapp deployed on HuggingFace spaces similar to other webapps such as https://hf.space/simonduerr/DiffDock and  https://hf.space/simonduerr/ProteinMPNN

Recording: https://youtu.be/J27XgQavGaU

Website: https://simonduerr.eu/