Uncovering viral glycoprotein conformational dynamics for rational vaccine design
Structure-based prediction and design have made tremendous progress over the last thirty years with Rosetta and, especially, since 2021, when AlphaFold2 was released. This progress to achieve the holy grail of computational structural biology was rewarded with the Nobel Prize for these two algorithms in 2024. The accurate prediction of the protein structure is the first step towards an in silico first strategy to design vaccines and antibodies. Structure-based modelling will also provide estimates for host-receptor interactions, such as antibody-antigen complexes. With this information, the missing link between sequence and function can be filled. Here, we are developing methods to predict structures of viral glycoproteins that will allow us to assess emerging viral variants. With structure-based methods, we will investigate the impact of the observed mutations in viral glycoproteins on the structure and the respective conformational states. This is critical for understanding the conformational space that mediates the fusion process. One major step to overcome here is the lack of predictive power of AI tools for structure prediction to provide pre- and post-fusion receptor states. We will subsequently predict the effect these mutations have on the structure. Together with the consortium partners, we will investigate whether our methods predict the effect and the respective function of the viral glycoprotein (variants). The major virus-host interaction we will study is the interaction with the immune system, with the spike protein being the major target of the humoral response to SARS-CoV-2.
Antibodies, as major determinants of this immune response, are useful research tools and therapeutics but are challenging for structure prediction and design. Moreover, antigen-antibody interactions are inherently hard to predict and evaluate, even with emerging AI tools, due to the lack of data and the complexity of the molecular interaction. Here, we will overcome these limitations and develop a new tool termed ANNtibody that takes atomic and electron density calculations into account. Thus far, these methods could not be employed for systems with more than a couple dozen atoms, but the training of AI on electron density data or on Density Functional Theory (DFT) calculations circumvents these resource limited steps. Data from NFDI4Chem and its associated repositories will be used to benchmark these.
With this increase in resolution, we hypothesise that our method will capture the complex interaction network in the antibody-antigen interface more accurately. These calculations will be used to predict antibody- antigen interactions, which we will challenge with the experimental design of antibodies for emerging SARS- CoV-2 variants using antibody interactions with the highly variable receptor-binding domain. With our method, we will update these antibody sequences and test experimentally whether we can overcome viral es- cape. Subsequently, we will design epitope focused immunogens based on SARS-CoV-2 epitopes that will elicit broadly neutralising antibody populations. All experimentally obtained data will be used to refine the developed methods. Data provided by C02 (Deckert/Deinhardt/Emmer) on virus-host interactions, by A02 (Friedel/Kühnert) and NFDI4Microbiota on viral sequences, and by B04 (Dittrich/McHardy) on surveillance will be essential for training and optimising our tool. These datasets will be integrated with our structural features, enabling our tool to generate predictions that will directly inform and support our partner projects. Altogether, we will generate computational structure-based tools that will help answers goals G1 and G3, providing insight into the host-receptor interaction. These tools will allow us to rapidly fight emerging viral infections and tune these tools for vaccine design, probing our ability to generate a broadly neutralising vaccine.
- WP 1: Prediction of stability and conformational landscape of viral glycoprotein with artificial-intelligence tools taking antigenic drift into consideration (Meiler, Schoeder)
- WP 2: Epitope prediction from interaction fingerprints using electron density resolution (Meiler)
- WP 3: (Re-) Design of antibodies for the SARS-CoV-2 spike RBD to overcome antigenic drift (Schoeder, Meiler)
- WP 4: Epitope-focusing and presentation of predicted and pan-coronavirus epitopes (Schoeder)
Team Members
N. N.
Doctoral Researcher
N. N.
Doctoral Researcher
Project-Specific Publications
2025
Self-supervised machine learning methods for protein design improve sampling but not the identification of high-fitness variants. Journal Article
In: Science advances, vol. 11, iss. 7, pp. eadr7338, 2025.
2024
Combining Rosetta sequence design with protein language model predictions using evolutionary scale modeling (ESM) as restraint. Journal Article
In: ACS synthetic biology, vol. 13, iss. 4, no. 4, pp. 1085–1092, 2024.
Combining machine learning with structure-based protein design to predict and engineer post-translational modifications of proteins. Journal Article
In: PLoS Comput Biol, vol. 20, iss. 3, no. 3, pp. e1011939, 2024.
2023
Modeling conformational states of proteins with AlphaFold Journal Article
In: Curr Opin Struct Biol, vol. 81, pp. 102645, 2023.
Biasing AlphaFold2 to predict GPCRs and kinases with user-defined functional or structural properties Journal Article
In: Front Mol Biosci, vol. 10, pp. 1121962, 2023.
2022
Epitope-focused immunogen design based on the ebolavirus glycoprotein HR2-MPER region Journal Article
In: PLoS Pathog, vol. 18, iss. 5, no. 5, pp. 1-29, 2022.
Sampling alternative conformational states of transporters and receptors with AlphaFold2 Journal Article
In: eLife, vol. 11, pp. e75751, 2022.
2021
Modeling immunity with Rosetta: Methods for antibody and antigen design Journal Article
In: Biochemistry, vol. 60, iss. 11, pp. 825–846, 2021.
2020
Predicting susceptibility to SARS-CoV-2 infection based on structural differences in ACE2 across species Journal Article
In: The FASEB Journal, vol. 34, iss. 12, no. 12, pp. 15946-15960, 2020.
Multi-state design of flexible proteins predicts sequences optimal for conformational change Journal Article
In: PLoS Comput Biol, vol. 16, iss. 2, pp. e1007339, 2020.