MIT presents Boltz-1, an open model for predicting 3D biomolecule structures

The scientific community worldwide has received an extraordinary opportunity for advancement in the field of biomolecular research thanks to a new tool developed at the Massachusetts Institute of Technology (MIT). Boltz-1, a revolutionary open-source artificial intelligence model, represents a key step towards the democratization of advanced technologies in biomedical research and drug development. This model allows for precise prediction of the three-dimensional structures of biomolecules such as proteins, opening doors for numerous scientific discoveries.

Open Innovation for a Better Tomorrow

Unlike similar solutions limited to academic purposes, such as AlphaFold3, Boltz-1 is fully open and available to all researchers, regardless of their institutional affiliation. The development of Boltz-1 was led by students Jeremy Wohlwend and Gabriele Corso from the Jameel Clinic for Machine Learning in Health at MIT, with support from scientists such as Sara Passar and professors Regina Barzilay and Tommi Jaakkola.

"Boltz-1 is not the end of the story," emphasized Corso during the presentation of the tool held on December 5 at MIT's Stata Center. "We aim to encourage global collaboration and enable the community to further develop this tool."

Technological Achievements That Change the Rules of the Game

Accurate prediction of protein and other biomolecule structures has been a key challenge in molecular biology for decades. The shape of a protein determines its function, and understanding these structures is crucial for developing new drugs and engineering proteins with specific functions. While AlphaFold2 from DeepMind set new standards, Boltz-1 further improved the technology by integrating innovative algorithms to enhance the efficiency of predictions and data usage.

One of the greatest challenges was working with the Protein Data Bank, a database containing biomolecular structures collected over the last 70 years. "There were many sleepless nights while we tackled issues related to inconsistent and heterogeneous data," recalled Wohlwend. Despite these challenges, Boltz-1 now achieves results comparable to AlphaFold3 in prediction accuracy.

Wide Support and Opportunities

The Boltz-1 project is supported by numerous organizations, including the National Science Foundation of the United States, the Jameel Clinic, and various research initiatives. The tool is already available on GitHub, where researchers can access the entire code and customize it to the specific needs of their research.

Further refinement of the model is planned to accelerate prediction times and expand its application possibilities. "This is just the beginning," added Professor Barzilay. "There are many potential directions for further development, and the community plays a crucial role in that process."

Global Impact

Experts from various fields have recognized the importance of this open model. "Boltz-1 will transform biomolecular research by enabling the availability of advanced tools to a broader community," said Mathai Mammen, president of Parabilis Medicines. Jonathan Weissman, professor of biology at MIT, emphasized how he expects a range of creative applications due to the openness of the model.

This tool not only accelerates the process of developing new drugs but also enables scientists to create customized biomolecular models for research in various fields, from genetics to oncology. Its impact on science has already been recognized as a crucial step toward a more inclusive and collaborative future in molecular sciences.