Skip to content

1. Introducing this course

Welcome to this VIB training. The authors of this course designed a full teaching day that combines theory with hands-on activities, allowing participants to apply the concepts in practice.

Section questions

  1. Why do we need this training ?

We have the structures now — what Next?

Where should we focus to understand protein behavior?

Proteins are often presented as static 3D structures, but in reality they are dynamic, context-dependent systems. Function does not emerge from a single snapshot—it arises from conformational landscapes, biophysical properties, and regulatory modifications that change across conditions, cell types, and time.

Post-translational modifications, such as phosphorylation, can reweight conformational states, alter interaction interfaces, and rewire signaling pathways without changing the underlying sequence. Likewise, mutations—whether disease-causing or benign—do not act uniformly: their impact depends on structural context, local dynamics, intrinsic disorder, and network connectivity within the protein.

This training bridges the gap between sequence, structure, and function by moving beyond static models to:

  • interpret backbone dynamics, intrinsic disorder, and folding propensities
  • place PTMs and mutations in their structural and biophysical context
  • understand how proteins sample multiple functional states, rather than a single “native” structure

By integrating structural models, biophysical annotations, and experimental proteomics evidence, we gain a mechanistic view of protein function, one that is essential for interpreting signaling, disease mutations, and regulatory complexity in real biological systems.

1.1 Introducing the training methodology

The protein of interest for this training is the human proto-oncogene tyrosine-protein kinase receptor RET, encoded by the RET gene. This protein is a receptor tyrosine-protein kinase involved in numerous cellular mechanisms including cell proliferation, neuronal navigation, cell migration, and cell differentiation. It exits is multiple conformational states (phospho and non-phospho). You will learn more about it’s biological context in Chapter 2.

  1. We will predict biophysical features for this protein using the Bio2Byte online platform, specifically the B2BTools suite. To enable a deeper analysis of these biophysical properties, we will generate a multiple sequence alignment (MSA) of sequences sharing at least 90% identity. This alignment will be used to identify conserved and variable patterns across homologous sequences.
  2. The MSA will be created using Clustal Omega, and the aligned kinase domain will be extracted using a Google Colab notebook.
  3. Post-translational modifications (PTMs) for the protein of interest will be explored using the Scop3P online platform, which is directly linked from the B2BTools prediction results. After analyzing the available information on modifications, structures, and experimental evidence, we will follow a more detailed protocol to link biophysical patterns with PTMs.
  4. Finally, the course will address the impact of mutations. We will show how to modify the wild-type sequence and assess the effect of single amino acid substitutions on biophysical profiles and predicted protein structures using AlphaFold v3.

Training material: Jupyter notebooks

Use the following link to launch the JupyterLab environment for this training:

Launch Binder

The environment may take a few minutes to build and start. Please open the link now so that it is ready when needed later in the training.

User-interface example

This is how the interface will look like:

ELIXIR Belgium
JupyterLab interface.

1.2 ELIXIR Belgium node services used in this training

Belgium is part of the ELIXIR Europe network as a National Node. The Belgian node, ELIXIR Belgium, provides both federal-level services and local initiatives, including research infrastructure, domain-specific services, training activities, and workshops.

ELIXIR Belgium
ELIXIR Belgium

This course focuses on two bioinformatics tools and resources: B2BTools, used to predict and analyse protein biophysical features, and Scop3P, used to explore post-translational modifications (PTMs).

1.2.1 Introduction to Bio2Byte Tools

DynaMine 1 is a predictor specifically designed to estimate protein backbone dynamics. Backbone dynamics are related to, but not the same as, protein disorder. DynaMine was trained using values derived from NMR chemical shift data and therefore captures protein movements in solution. Its training set includes both fully folded proteins and intrinsically disordered proteins.

The B2BTools tool suite 2 extends the original DynaMine predictor by including several predictors developed by the Bio2Byte lab.

In addition to backbone dynamics, it provides predictions for side-chain dynamics and conformational preferences (alpha helix, beta sheet, and coil), all derived from NMR data and trained using the same methodology. The platform also includes predictors for early folding regions (EFoldMine) 3, beta-sheet aggregation (AgMata) 4, and protein disorder (DisoMine) 5.

1.2.2 Introduction to Scop3P (and Scop3PTM)

Scop3P 6, developed at Ghent University and available online since June 2019, is a dedicated resource to explore and interpret the impact of phosphorylation sites on human protein structure and function. It supports researchers in analysing individual phosphosites or phosphoproteins within a structural, biophysical, and biological context.

The resource integrates public data from several major international databases, including UniProtKB and the Protein Data Bank (PDB). In addition, it incorporates reprocessed mass spectrometry-based phosphoproteomics data from PRIDE/ProteomeXchange. These datasets are collected worldwide, making Scop3P a strongly international and community-driven resource.

About the future of Scop3P and the development of Scop3PTM

Scop3P is being extended to Scop3PTM

  • Scop3PTM integrates information from different knowledge bases and shows how re-analysis of large scale public proteomics data sets can add an additional level of significance and confidence to the PTM-sites.
  • Scop3PTM system will provide a unique and powerful resource to understand the impact of PTM-sites on human protein structure-function relationship.

Beta access is available at https://iomics.ugent.be/scop3ptm/.

Let’s get started: Go to chapter 2

Next chapter explains the biological context.

1.3 Rethinking the protein structure and function

From “Assessing the relation between protein phosphorylation, alphafold3 models, and conformational variability” 7:

Proteins perform diverse functions critical to cellular processes. Transitions between functional states are often regulated by post-translational modifications (PTMs) such as phosphorylation, which dynamically influence protein structure, function, folding, and interactions. Dysregulation of PTMs can therefore contribute to diseases such as cancer and Alzheimer’s. However, the structure–function relationship between proteins and their modifications remains poorly understood due to a lack of experimental structural data, the inherent diversity of PTMs, and the dynamic nature of proteins.

Recent advances in deep learning, particularly AlphaFold, have transformed protein structure prediction with near-experimental accuracy. However, it remains unclear whether these models can effectively capture PTM-driven conformational changes, such as those induced by phosphorylation. Here, we systematically evaluated AlphaFold models (AF2, AF3-non phospho, and AF3-phospho) to assess their ability to predict phosphorylation-induced structural diversity. By analyzing experimentally derived conformational ensembles, we found that all models predominantly aligned with dominant structural states, often failing to capture phosphorylation-specific conformations. Despite its phosphorylation-aware design, AF3-phospho predictions provided only modest improvement over AF2 and AF3-non phospho predictions.

Our findings highlight key challenges in modeling PTM-driven structural landscapes and underscore the need for more adaptable structure prediction frameworks capable of capturing modification-induced conformational variability.

1.3.1 Proteins are dynamic systems, not static objects

Image on elixir landing page

1.3.2 Limitations of deep learning–based structure prediction

Image on elixir landing page
Image on elixir landing page

1.3.3 Memorization and dominant conformations

Image on elixir landing page

  1. Elisa Cilia, Rita Pancsa, Peter Tompa, Tom Lenaerts, and Wim F Vranken. From protein sequence to dynamics and disorder with dynamine. Nature communications, 4:2741, 2013. URL: https://doi.org/10.1038/ncomms3741, doi:10.1038/ncomms3741. ↩

  2. Jose Gavalda-Garcia, Adrián Díaz, and Wim Vranken. Bio2byte tools deployment as a python package and galaxy tool to predict protein biophysical properties. Bioinformatics (Oxford, England), 40(9):btae543, September 2024. URL: https://europepmc.org/articles/PMC11873786, doi:10.1093/bioinformatics/btae543. ↩

  3. Daniele Raimondi, Gabriele Orlando, Rita Pancsa, Taushif Khan, and Wim F Vranken. Exploring the sequence-based prediction of folding initiation sites in proteins. Scientific reports, 7(1):8826, August 2017. URL: https://europepmc.org/articles/PMC5562875, doi:10.1038/s41598-017-08366-3. ↩

  4. Gabriele Orlando, Alexandra Silva, Sandra Macedo-Ribeiro, Daniele Raimondi, and Wim Vranken. Accurate prediction of protein beta-aggregation with generalized statistical potentials. Bioinformatics, 36(7):2076–2081, 12 2019. URL: https://doi.org/10.1093/bioinformatics/btz912, arXiv:https://academic.oup.com/bioinformatics/article-pdf/36/7/2076/50670159/bioinformatics_36_7_2076.pdf, doi:10.1093/bioinformatics/btz912. ↩

  5. Gabriele Orlando, Daniele Raimondi, Francesco Codicè, Francesco Tabaro, and Wim Vranken. Prediction of disordered regions in proteins with recurrent neural networks and protein dynamics. Journal of Molecular Biology, 434(12):167579, 2022. URL: https://www.sciencedirect.com/science/article/pii/S0022283622001590, doi:https://doi.org/10.1016/j.jmb.2022.167579. ↩

  6. Pathmanaban Ramasamy, Demet Turan, Natalia Tichshenko, Niels Hulstaert, Elien Vandermarliere, Wim Vranken, and Lennart Martens. Scop3p: a comprehensive resource of human phosphosites within their full context. Journal of proteome research, 19(8):3478—3486, August 2020. URL: https://doi.org/10.1021/acs.jproteome.0c00306, doi:10.1021/acs.jproteome.0c00306. ↩

  7. Pathmanaban Ramasamy, Jasper Zuallaert, Lennart Martens, and Wim F Vranken. Assessing the relation between protein phosphorylation, alphafold3 models, and conformational variability. Protein science : a publication of the Protein Society, 35(1):e70376, January 2026. URL: https://europepmc.org/articles/PMC12723721, doi:10.1002/pro.70376. ↩