Mapping and Learning Protein Structural Diversity Across Random Sequence Space
EPFL BIOE TALKS SERIES (sandwiches provided)
Abstract:
Biological proteins represent only a few tightly clustered families within the vast sea of amino acid combinations. While nature has explored only a fraction of this landscape, the 'dark protein space' offers untapped potential for de novo design and the study of gene birth.
Through high-throughput structural screening of large random-sequence libraries, we show that this space is surprisingly heterogeneous. Beyond disordered or toxic aggregates, we identified “benign” compact architectures that mimic globular proteins and bypass cellular chaperone defenses. By training a neural network on these data, we demonstrate that protein structural potential is predictable and generalizes across natural proteomes. This suggests that biology-like folds are frequent in random space, offering a path to expand generative design far beyond evolutionary priors.
Bio:
Klára Hlouchová studied chemistry at Charles University in Prague and was trained as a biochemist at IOCB (PhD in 2009). Her postdoctoral research in evolution of metabolic pathways with Prof. Shelley Copley at University of Colorado Boulder (2011-2013) sparked her curiosity in life’s early origins and the divide between non-viable and viable, converging with a long-standing interest in astrobiology. She started her independent research at Charles University (Department of Biochemistry and later Department of Cell Biology) in 2016. Her group of Synthetic Biology explores fundamental questions about the origin and evolution of proteins, be it in the era of prebiotic Earth, extant biological or synthetic life. They also have several projects focusing on what it takes for cells to emerge, with the aid of proteins and other key life’s molecules all playing along. Her research has been supported by standard Czech and European funding schemes but also by several collaborative and interdisciplinary grants, such as HFSP and VW Stiftung.
Zoom link (with one-time registration for the whole series) for attending remotely: https://go.epfl.ch/EPFLBioETalks
Instructions for 1st-year Ph.D. students planning to attend this talk, who are under EDBB’s mandatory seminar attendance rule:
IN CASE you cannot attend in-person in the room, please make sure to
Abstract:
Biological proteins represent only a few tightly clustered families within the vast sea of amino acid combinations. While nature has explored only a fraction of this landscape, the 'dark protein space' offers untapped potential for de novo design and the study of gene birth.
Through high-throughput structural screening of large random-sequence libraries, we show that this space is surprisingly heterogeneous. Beyond disordered or toxic aggregates, we identified “benign” compact architectures that mimic globular proteins and bypass cellular chaperone defenses. By training a neural network on these data, we demonstrate that protein structural potential is predictable and generalizes across natural proteomes. This suggests that biology-like folds are frequent in random space, offering a path to expand generative design far beyond evolutionary priors.
Bio:
Klára Hlouchová studied chemistry at Charles University in Prague and was trained as a biochemist at IOCB (PhD in 2009). Her postdoctoral research in evolution of metabolic pathways with Prof. Shelley Copley at University of Colorado Boulder (2011-2013) sparked her curiosity in life’s early origins and the divide between non-viable and viable, converging with a long-standing interest in astrobiology. She started her independent research at Charles University (Department of Biochemistry and later Department of Cell Biology) in 2016. Her group of Synthetic Biology explores fundamental questions about the origin and evolution of proteins, be it in the era of prebiotic Earth, extant biological or synthetic life. They also have several projects focusing on what it takes for cells to emerge, with the aid of proteins and other key life’s molecules all playing along. Her research has been supported by standard Czech and European funding schemes but also by several collaborative and interdisciplinary grants, such as HFSP and VW Stiftung.
Zoom link (with one-time registration for the whole series) for attending remotely: https://go.epfl.ch/EPFLBioETalks
Instructions for 1st-year Ph.D. students planning to attend this talk, who are under EDBB’s mandatory seminar attendance rule:
IN CASE you cannot attend in-person in the room, please make sure to
- send Fiorella Ghisays a note well ahead of time (ideally before seminar day), informing that you plan to attend the talk online, and, during seminar:
- be signed in on Zoom with a recognizable user name (not any alias making it difficult or impossible to identify you).
Practical information
- Informed public
- Registration required
Organizer
- Prof. Anne-Florence Bitbol, Institute of Bioengineering
Contact
- Fiorella Ghisays, Institute of Bioengineering (IBI)