Computational structural biology workshop in six takeaways
AI takes centre stage at an EMBO workshop exploring emerging trends, collaborative opportunities, and recent advances in computational structural biology
By Suhad Al-Salihi, Senior Lecturer, University of Technology, Baghdad
The microscopic world of proteins, DNA, and other biomolecules is constantly in motion. Understanding these dynamics is key to advancing medicine and revealing how cells function. From 2–5 December, the EMBO Computational Structural Biology workshop brought together global researchers to showcase the latest tools in structure prediction, protein design, drug discovery, and molecular interaction modelling. Here are six key takeaways from the workshop:
AI remains a driving force for structural prediction innovation.
Speakers showed that AlphaFold3 can model much more than just neatly folded proteins. It can also handle sugars attached to proteins, custom small molecules, interactions with DNA and RNA, antibody-antigen binding, and flexible or disordered protein regions. They shared practical tips, such as how to spot uneven confidence plots, why short molecular dynamics relaxation can improve models, and how designing the right protein constructs helps when studying DNA or RNA binding.
Researchers compared AlphaFold2 and AlphaFold3 in real biological systems and emerging tools for more reliable model selection. One speaker encouraged the research community to contribute more data to shared databases to further improve AI models, particularly for unexplored ligands and polymer contexts.
Structural dynamics are key to understanding function – bacterial surfaces and beyond.
Simulations of bacterial membranes, viral proteins, and cell-scale assemblies showed why structure alone is only half the story. The speakers explained how to scale up molecular dynamics simulations, compare them to fast atomic force microscopy data, including realistic details such as the natural chemistry of bacterial membranes. Overall, this work is moving toward more realistic, dynamic models of how molecules behave.
We’re entering an era of evolution-guided protein design.
Several talks demonstrated how natural structural principles are now driving generative design. Researchers explained that many protein shapes emerge from expandable ancestral building blocks and that models like ESMFold can help validate designs when multiple sequence alignments are inadequate. TIM barrel protein folds and novel catalytic scaffolds (artificially designed protein structures often based on TIM barrel folds) were hailed as successful examples of how AI-assisted design is increasingly matching what works in real lab experiments.
Deep learning underpins structural bioinformatics across experiment and computation.
Cryo-EM, tomography, and crosslinking now routinely combine with computer-based structural predictions. Several researchers showed how crosslinking data helps improve models of large protein complexes, how small molecules can be checked against experimental density maps, and how whole-cell cryo-electron tomography is getting better but still works best for large protein assemblies. It was clear that experimental and computational research are now moving forward together rather than advancing in isolation.
Structure-based drug discovery is being reshaped by massive chemical space exploration and industrial data.
Rapid virtual screening, AI that can design new (or optimise old) antibodies, and physics-based binding predictions are changing how medicines are made. Researchers can now test billions of molecules computationally, using detailed simulations to realistically predict how they’ll behave when lab work begins. The goal is to make early drug development decisions faster and better informed.
Scalable infrastructure is essential for future breakthroughs.
Industrial-scale computing and flexible modelling platforms like HADDOCK3 are becoming essential to handle the growing complexity and scale of modern computational modelling tasks. Together, these advances make high-level structural biology more accessible to researchers everywhere.
The workshop was ultimately more than a series of talks; it shared how structural biology is heading into the next decade. The mix of creativity, technology, and collaboration made it clear how important it is for scientists to connect in person, exchange experiences, and shape the direction of the field together. In addition to the incredible knowledge gained, events like this provide inspiration that comes from being part of a community pushing the boundaries of what we can understand and engineer in life – a gratifying experience for any researcher in computational biology, structural biochemistry, biophysics, or drug discovery.
