AI and Structural Biology

Artificial Intelligence (AI) is playing an ever-increasing role in our lives. In science, and structural biology, it is no different - from prediction software, such as AlphaFold, to machine learning (ML) processes, the continuing fusion of AI with experimental structural characterisation and project design.

As a structural biology research infrastructure, Instruct-ERIC provides expertise on how to get the best out of AI and structural biology. How to get the most out of prediction tools, when best to use them, and how to validate the predictions to make sure they are accurate and provide enough biological context.

What is AlphaFold?

AlphaFold is a structure prediction tool, using the amino acid sequence to predict the 3D structure of proteins. Launched in 2021 by Google DeepMind and EMBL-EBI, AlphaFold initially won plaudits for its exceptional accuracy over other prediction services. Since then, there have been several updates to the system - currently the AlphaFold server is powered by AlphaFold 3.

As well as proteins, AlphaFold 3 can generate accurate structures containing DNA, RNA, ligands, ions, and can also model chemical modifications for proteins and nucleic acids. These updates ensure that the system can provide more biological context to a structure prediction - providing an insight into how the protein might function too.

How to use AlphaFold

For those who are new to AlphaFold, EMBL-EBI provides an online walk-through course to get up to speed. This outlines:

• How AlphaFold works and its strengths and limitations
• How AlphaFold predictions were validated experimentally
• The fundamental concepts behind AlphaFold and why it is considered a significant breakthrough in protein structure prediction
• The best way to predict protein structures
• The best way to access pre-computed predictions from the AlphaFold Protein Structure Database
• How to evaluate predicted structures from AlphaFold by integrating the different confidence metrics

See the online course from EMBL-EBI here.

Using AlphaFold for Structural Biology

The important thing to remember about structural prediction tools is that they are just that: a prediction.

Instruct-ERIC provides access to structural biology services across Europe for experimental structural characterisation. Our centres are filled with structural biology experts, who can provide their experience and expertise on how best to blend prediction with experimentation. They can also offer their expertise with prediction programmes developed by Instruct facilities, including the HADDOCK docking prediction tool, and AlphaFill to gather even more context about predicted structures.

• How to integrate prediction tools into a structural biology project
• What can be achieved with experimentation rather than prediction
• Which techniques are most suitable for structure validation/characterisation

Speak to our experts to get these questions answered.

View all Validation Services           Contact our Experts

An open source version of AlphaFold3 is currently being developed, to find out more and try the initial release of HelixFold3 visit here.

Any other AI structural biology services to include here? Let us know!

Machine Learning in Structural Biology

In addition to structure prediction, AI has some innovative functions within structural biology. As part of the Fragment-Screen project, Instruct is working with industry experts such as IBM Research to develop ML systems that can identify potential targets for fragment-based drug development (FBDD).

With ML, researchers can narrow down thousands of targets to a select few. This is another example of how AI can help design the experiments which are then performed in the lab to test their accuracy. Example papers below outline the potential use of AI in identifying potential pharmaceuticals for FBDD, narrowing down the list and saving months in the initial experimental design phase.

Using AI Generative Modelling to Identify Novel SARS-CoV-2 Drug Targets at Instruct-UK

In recent years, the field of structural biology has undergone a revolution thanks to the exceptional predicting capabilities of AI models, such as AlphaFold. With these tools, predicting the 3D structure of proteins has become quicker, and more accurate...

Find out more

The Blending of Experimental and Curiosity-Driven Machine Learning in Fragment-Screen

Coordinated by Instruct-ERIC, the Fragment-Screen project aims to develop innovative instrumentation and make significant advances in fragment-based drug discovery (FBDD). The project brings together experts in a variety of structural biology domains...

Find out more

AI Projects

Instruct is part of several projects with AI as a central theme, to develop new tools and techniques in structural biology.

AI4Life

AI4LIFE aims to build bridges between the life science community and the machine learning/artificial intelligence community. AI has enormous potential for advancing life sciences - AI4LIFE is building an open, accessible, community-driven repository of FAIR pre-trained AI models and develop services to deliver these models to life scientists - read more here. The BioImage Model Zoo has several AI-driven models to gain better insights from biological images.

Fragment-Screen

Coordinated by Instruct-ERIC, the Fragment-Screen project aims to develop innovative instrumentation, workflows and experimental and computational methodologies to accelerate the development of new pharmaceuticals using the approach of fragment-based drug discovery (FBDD). It will improve current bottlenecks in signal-to-noise (NMR), automation of data analysis (X-ray), and general applicability of cryo-EM for high-throughput - with a strong collaboration with AI development teams and industrial partners. Read more here.