GAIL Seminar Series: Protein inverse-folding - a simple idea with a vast potential Generative AI is rapidly transforming the field of protein design, increasingly making it possible to create novel, complex proteins from scratch. Several different types of machine learning models have contributed to these developments, but one stands out both for its simplicity and ubiquitous use in current design pipelines: the inverse folding model. In his GAIL Seminar, Professor Boomsma will give a brief introduction to this modeling task, how it represents one of the early examples of deep learning in protein space, and how it foreshadowed the current practice of pre-trained foundation models. He will then discuss some more recent developments in the field, and the impact of artificial training data in the form of predicted structures. Finally, he will present some recent work from my lab on future perspectives of these models, focusing on the question of why likelihoods from inverse folding models correlate so closely with protein stability. Biography Wouter Boomsma is a professor at the Department of Computer Science, University of Copenhagen, where he leads a group on Machine Learning in Biology. His team is part of the MLLS Center for Basic Machine Learning Research in Life Science. His current research interests are focused on probabilistic machine learning methods for understanding sequence/structure/function relationships in proteins and DNA. He has held postdoctoral positions in various universities, including Cambridge with Michele Vendruscolo, Lund with Anders Irback, DTU with Jesper Ferkinghoff-Borg and Copenhagen with Kresten Lindorff-Larsen. He did his PhD at the University of Copenhagen with Thomas Hamelryck and Anders Krogh. He is originally trained as a computer scientist from the University of Aarhus. Apr 28 2025 14.00 - 15.30 GAIL Seminar Series: Protein inverse-folding - a simple idea with a vast potential The next GAIL Seminar will be delivered by Wouter Boomsma from University of Copenhagen where he will discuss how Generative AI is helping to rapidly transforming the field of protein design. Room 2.35, Edinburgh Futures Institute Register
GAIL Seminar Series: Protein inverse-folding - a simple idea with a vast potential Generative AI is rapidly transforming the field of protein design, increasingly making it possible to create novel, complex proteins from scratch. Several different types of machine learning models have contributed to these developments, but one stands out both for its simplicity and ubiquitous use in current design pipelines: the inverse folding model. In his GAIL Seminar, Professor Boomsma will give a brief introduction to this modeling task, how it represents one of the early examples of deep learning in protein space, and how it foreshadowed the current practice of pre-trained foundation models. He will then discuss some more recent developments in the field, and the impact of artificial training data in the form of predicted structures. Finally, he will present some recent work from my lab on future perspectives of these models, focusing on the question of why likelihoods from inverse folding models correlate so closely with protein stability. Biography Wouter Boomsma is a professor at the Department of Computer Science, University of Copenhagen, where he leads a group on Machine Learning in Biology. His team is part of the MLLS Center for Basic Machine Learning Research in Life Science. His current research interests are focused on probabilistic machine learning methods for understanding sequence/structure/function relationships in proteins and DNA. He has held postdoctoral positions in various universities, including Cambridge with Michele Vendruscolo, Lund with Anders Irback, DTU with Jesper Ferkinghoff-Borg and Copenhagen with Kresten Lindorff-Larsen. He did his PhD at the University of Copenhagen with Thomas Hamelryck and Anders Krogh. He is originally trained as a computer scientist from the University of Aarhus. Apr 28 2025 14.00 - 15.30 GAIL Seminar Series: Protein inverse-folding - a simple idea with a vast potential The next GAIL Seminar will be delivered by Wouter Boomsma from University of Copenhagen where he will discuss how Generative AI is helping to rapidly transforming the field of protein design. Room 2.35, Edinburgh Futures Institute Register
Apr 28 2025 14.00 - 15.30 GAIL Seminar Series: Protein inverse-folding - a simple idea with a vast potential The next GAIL Seminar will be delivered by Wouter Boomsma from University of Copenhagen where he will discuss how Generative AI is helping to rapidly transforming the field of protein design.
