Best poster prizes at the EMBO Workshop ‘Computational structural biology’ – Course and Conference Office


Our mission is to train scientists. This blog is a platform for us to share updates on our annual programme, tips and tricks for scientists, new e-learning opportunities, and sometimes just something to make you smile.

Best poster prizes at the EMBO Workshop ‘Computational structural biology’

At the beginning of December 2023, the EMBO Workshop ‘Computational structural biology’ brought together scientists to discuss the manifold trials to apply artificial intelligence (AI) techniques, such as deep learning, to solve great challenges in the field of structural biology biophysics of biomolecules.

The new breakthrough in this field is called AlphaFold, and, during the conference, computational and experimental structural biologists and biophysicists discussed the potential pros and cons of AI applications in structural biology and bioinformatics.

Today, we are proud to announce the winners of the best poster prizes, recognizing outstanding contributions to the field. Take a moment to explore the exciting research articles and posters of two of them and get inspired by those remarkable winners!

Computational modeling of OR51E2 via de novo structure prediction and molecular dynamics simulation

Presenter: Riccardo Capelli, University of Milan, Italy

Collaborators: Mercedes Alfonso Prieto, Forschungszentrum Jülich, Germany

Riccardo Capelli

Abstract: Understanding olfactory receptors (ORs) at the atomistic level presents a remarkable challenge, primarily due to the complexities associated with experimental and computational approaches in determining and predicting the structures of this diverse family of G-protein coupled receptors (GPCRs).

In this study, we introduce a novel protocol designed to overcome these hurdles. Our approach involves a series of molecular dynamics simulations, utilizing as a starting point structures generated de novo through state-of-the-art machine learning algorithms. We applied this protocol to investigate the well characterized human OR51E2 receptor, shedding light on the critical role of simulations in refining and assessing computational models.

A key result from our research is the significance of a sodium ion within a binding site close to residues D2.50 and E3.39 in stabilizing the inactive state of the receptor. Notably, we observe the conservation of these two acidic residues across the extensive repertoire of human ORs, suggesting that the necessity for a sodium ion extends to the approximately 400 other members of this receptor family. This insight suggests the broader implications of our findings.

Finally, our work coincides with the publication of a CryoEM structure of the same receptor in its active state, the first experimental conformation for a OR ever solved. This reinforces the complementary nature of our in-silico protocol within the growing field of OR structural elucidation. By offering a computational tool to augment experimental techniques, we contribute to the collective efforts aimed at unraveling the complexity of ORs and their vital role in olfaction.

Poster not available

Aggregating and optimising structure predictions with AlphaFold2

Presenter: Jannik Adrian Gut, University of Bern, Switzerland

Collaborators: Thomas Lemmin, University of Bern, Switzerland

Jannik Adrian Gut

Abstract: Protein folding has experienced a transformative revolution with the introduction of AlphaFold2, establishing itself as the gold standard for many protein folding tasks.

Nonetheless, AlphaFold2 is not a universal solution as its performance is intricately tied to the quality of the provided multiple sequence alignment (MSA). Therefore, many alternative folding models with distinct properties, specialisations and biases have emerged to address these limitations.

A prevailing hypothesis suggests that in the case of AlphaFold2, the MSA primarily serves to delineate the correct protein fold neighbourhood, while the subsequent components of the model define an learned energy function to refine the fold predictions independently of sequence data.

In this study, we investigated this hypothesis by exploring the capacity of AlphaFold2 to optimise different structure predictions through the incorporation of templates and diverse MSA depth. Our investigation encompasses monomers, oligomers, and select synthetic test cases. We show the strengths and limitations of various aggregation approaches from ensembles of structure predictions derived from different folding models. Our findings represent a step forwards in the interpretability of AlphaFold2 and offer new insights into the energy function hypothesis.

View poster

The EMBO Workshop ‘Computational structural biology’ took place from 6 – 9 December 2023 at EMBL Heidelberg and virtually.