The ability to accurately predict protein structure is critical for understanding how proteins work, how proteins interact with other molecules, and how they can be targeted by drugs. Until recently, accurately predicting protein structure was a challenging task, especially when there is little or no homology to known structures. It could take months to determine a protein structure using experimental methods such as x-ray crystallography, cryo-electron microscopy, or nuclear magnetic resonance spectroscopy, and it is difficult and costly to collect the amount of material to carry out the experiments.
Applications of AlphaFold and OpenFold for Drug Discovery
In recent years, deep learning has made significant progress in the field of protein folding. AlphaFold2, developed by DeepMind, is a deep learning-based protein structure prediction tool that has shown great promise in accurately predicting the 3D structure of proteins in a matter of minutes to hours. OpenFold, developed by the lab of Mohammed AlQuraishi at Columbia University, has a similar approach with a further optimized model inference stage. Both of these tools have the potential to revolutionize the way we study proteins and their interactions.
A growing field of research has been exploring how best to leverage models from AlphaFold and OpenFold. Recent work by Lyu et al and the Critical Assessment of Computational Hit-Finding Experiments have demonstrated that structural models from AlphaFold can be used effectively in virtual screening studies and the hit rate is very similar to when experimental structures are used. Outside of small molecule applications, AlphaFold has broad use in modeling protein-protein and protein-peptide complexes for biologic design workflows. For example, many studies have shown competitive performance of AlphaFold2 in predicting protein-protein or protein-peptide complexes, including the recent results from the 15th Community Wide Experiment on the Critical Assessment of Techniques for Protein Structure Prediction.
Overcoming the Barriers in Structure Prediction with Out-of-the-Box OpenFold and AlphaFold2 AI Models Available in Discovery Studio Simulation
We are excited to announce that Dassault Systèmes BIOVIA will offer OpenFold (for monomer structure prediction) and AlphaFold2 (for multimer structure prediction) as part of the BIOVIA Discovery Studio Simulation service on Cloud, allowing researchers, scientists, and organizations to easily access the powerful capabilities of these methods and accelerate their research.
With our cloud offering, researchers can easily access the power of these methods without the need for expensive hardware or technical expertise. Our cloud service provides a user-friendly interface through the Discovery Studio Simulation client, easily running prediction for one or more protein sequences in batch mode. The protocol also provides comprehensive output data and interactive plots for the analysis and assessment of the predicted structures that are familiar to many researchers.
Discovery Studio Simulation combines deep learning methods with physics-based approaches conveniently in a single application and provides easy workflows to navigate complex computational tasks. The application makes available a protocol to predict the 3D structures of target protein sequences with OpenFold or AlphaFold2. These models typically require further post-processing for downstream workflows such as docking or molecular simulation, and Discovery Studio Simulation offers multiple protocols to address this requirement including protein preparation by assigning protonation states, model refinement via energy minimization. Once a structural model is acquired, a number of physics-based protocols from Discovery Studio Simulation can be leveraged for diverse computational tasks including virtual screening via docking, sampling conformational states with enhanced sampling molecular dynamics, or studying protein formulation properties.
Predicting protein structures using AlphaFold2 and OpenFold is just the beginning for potential application of deep learning models. Many new opportunities are opening up, facilitating potential new AI methods for protein engineering and drug development.
Get started with our OpenFold and AlphaFold2 cloud service today, let these powerful methods drive new discoveries in your research in the area of drug development, biotechnology, and beyond.
Watch the video to learn how to analyze the results of the AlpahFold2/OpenFold predictions in Discovery Studio Simulation
Lisa Yan Dr. YAN has a Ph.D. in computational chemistry and worked on developing computational methods for drug discovery for many years. As the senior manager of the Discovery Studio R&D team, she has been a key contributor in designing the current macromolecule modeling packages in Discovery Studio Simulation and led the development of the scientific methods and user interface of the tools in this area, including homology modeling, sequence analysis, antibody modeling, and protein design.
Reed Harrison Dr. Reed HARRISON is a R&D software engineering specialist at BIOVIA, Dassault Systèmes. With a background in protein engineering, computational chemistry, and bioinformatics; a Ph.D. in bioengineering; and an interest in methods to probe and engineer the protein structure-function paradigm, Reed has worked to support protein modeling and simulation activities in BIOVIA.
Rohith MOHAN Rohith received his Ph.D. in Bioengineering from University of California, Riverside. As part of his doctorate research he studied the complement system and the molecular basis of immune-related diseases by using MD simulations for mechanistic studies, peptide design, and virtual drug screening. He joined BIOVIA in 2018 as part of the Discovery Studio R&D team for the development of modeling and simulation tools for drug design and discovery. He currently leads the Protein Modeling and Simulation team in Discovery Studio where he works on force field parameterization, free energy methods, simulation methods, and the intersection of AI/ML in these areas.