Why has AlphaFold Protein Structure Database become so important for Protein Misfolding
- Details
- Written by: Developer SM
- Category: Perspectives
- Hits: 626
![Image for Revolutionizing Treatment: Gain Therapeutics Leads the Way in AI-Driven Drug Discovery for Complex Diseases](/images/Why-has-AlphaFold-Protein-Structure-Database-become-so-important-for-Protein-Misfolding-1-1024x569.jpg)
Why has AlphaFold Protein Structure Database become so important for Protein Misfolding
Why is the AlphaFold DataBase exciting?
DeepMind created the artificial intelligence system, AlphaFold, to predict the three-dimensional (3D) structure of thousands of proteins and compiled these structures into the AlphaFold DataBase. Access to protein structures has impeded biological research for decades. Scientists have been working for over 60 years to experimentally determine the 3D structures of proteins and first started compiling structures in the 1970s, but the AlphaFold database has accelerated this work.
This database is remarkable, offering the predicted 3D protein structures for many of the proteins found in humans. In addition, protein structures are also available for an additional 20 model species that we commonly use in scientific and medical research. In its entirety, this database allows scientists to access hundreds of thousands of structural predictions.
How can scientists use the AlphaFold DataBase?
The AlphaFold DataBase creates great opportunities for structural biologists around the world. These scientists use 3D protein structures for basic biology to understand how the structure and function of a protein are related. They can use this data to explore how proteins interact with other proteins or molecules, slowly solving the molecular mysteries of how each cellular machine works. In many cases, structural data and experimental functional data are not available or happening simultaneously in many areas of biology. The 3D predictions will allow scientists to jumpstart many structural determination studies so that they can compare the AlphaFold prediction to the actual 3D protein structure.
In recent publications regarding AlphaFold, the authors said the AI system provided high-quality predictions for the structure of many human proteins that are relevant to therapeutic research. This innovative tech allows scientists to harness the predictive power of AlphaFold as protein structures are exceedingly important for the development of successful disease-modifying drugs. However, we are only beginning to understand what classes of disease are directly benefited by access to 3D protein structure predictions.
What benefits does the database provide for protein misfolding research?
Protein misfolding diseases are a class of disorders characterized by proteins that are folded into the incorrect shape. Misfolded proteins are notoriously difficult when it comes to experimentally solving a structure. Currently, scientists use expensive and often time-consuming techniques like cryo-electron microscopy (cryo-EM) to crack the structure of misfolded proteins. This technique involves flash-freezing a solution of a purified protein that is then hit with a beam of particles called electrons. A camera can detect this signal and uses this input to produce an atomic resolution image of the protein. While Cryo-EM is the gold standard, misfolded proteins usually have large intrinsically disordered and flexible regions that are difficult to solve. Similarly, when these unstructured regions are present in proteins, AlphaFold struggles to predict a high-confidence structure. The AI also has a long way to go before it has learned how to predict the effect of mutations on protein structure, a significant issue in protein misfolding disorders. However, there are many ways in which the AlphaFold DataBase is extremely useful for protein misfolding research.
Protein misfolding diseases are often caused by mutations in proteins or the chaperones that help them fold. This leads to loss of function and often the build-up of large aggregates of misfolded protein that are toxic to cells. Despite the AlphaFold algorithm struggling to predict high-resolution structures when proteins are intrinsically disordered, it is capable of predicting some kind of 3D structure for every imaginable amino acid sequence. Even if the structure is low-confidence, scientists can pair that data with other experimental techniques in hopes of arriving at a final 3D shape. It is even possible that this process is how the AlphaFold AI’s prediction accuracy is validated. By pairing these data together, scientists can use the structural predictions to design drugs or antibodies that target a specific shape of a protein.
When it comes to mutations in a protein, AlphaFold and its paired database can be used to answer multiple questions. Structural biologists can use what they know about bonds, or interactions, between amino acids to predict how they interact with each other. This allows them to take an available 3D prediction of a protein and hypothesize how a mutation could change the shape. This hypothesis can then be tested experimentally and by AlphaFold. Because the AI can theoretically provide a prediction for every amino acid sequence, it should be able to predict a general structure for a mutated amino acid sequence.
For some protein misfolding diseases, a drug may be designed to rescue the function of a misfolded protein. As the AI learns, its applications will expand, and it is possible that the AlphaFold DataBase will grow to include 3D predictions of proteins and their misfolded counterparts. This would make it possible to predict how a small molecule drug could bind and force a misfolded protein into its intended shape, restoring its function. It may even be possible to use the database to learn how we can prevent a protein from misfolding. In every scenario, the models made available by the AlphaFold DataBase can inform how therapeutics for protein misfolding diseases are selected during development.
How is Gain Therapeutics using the AlphaFold DataBase?
Gain Therapeutics has already started to use the public protein structures in the AlphaFold database for novel target discovery. Gain Therapeutics’ drug discovery platform uses 3D structures as the input. Traditionally this would be through well documented and validated structures, but due to the accuracy of alpha folds predictions as well as the variety of proteins in the AlphaFold DataBase, Gain Therapeutics now has access to more than double the protein targets. Target discovery work has begun and it is likely that Gain will discover and develop a novel drug based on information from the AlphaFold DataBase.
Latest Articles