Drug Discovery Institutes may be able to innovate more quickly with new available AI models.
This new model, called Boltz-2, is the successor to Boltz-1, a deep learning model like Alphafold that predicts biomolecule interactions. The main new product of Boltz-2 is a binding affinity prediction model that achieves results comparable to free energy perturbation (FEP) simulations, but runs 1,000 times faster.
Lab Management Certificate
Lab Management Certificates are more than training. This is a professional advantage.
Earn important skills that make a measurable difference and IACET-approved CEU.
Like its predecessor, Boltz-2 is open source licensed by the Massachusetts Institute of Technology (MIT) and has no restrictions on academic or commercial use. According to a press release, joint modeling of Boltz-2 complex structures and binding affinity makes “silico screening practical for early stage drug discovery.”
The new model promises faster and more efficient research
An important component of drug design is to assess how strongly a drug molecule binds to a target protein called its binding affinity. Binding affinity is typically predicted in FEP, a physics-based technology. Although accurate, FEP is time- and resource-intensive and can limit research advancements.
Boltz-2's combined affinity prediction engine can run 1000 times faster than FEP, while achieving comparable accuracy representing Biopharma Labs opportunities.
“The structure and affinity prediction of Boltz-2 allow for a very inexpensive, yet relatively accurate method of testing the compounds before ordering,” said Gabriele Corso, a graduate student at MIT and one of the project's lead developers. “moreover, [by] Fully leveraging the virtual screening pipeline built on top of Boltz-2 allows scientists to reduce the number [of] Experimental verification[s] We needed to find an effective molecule. ”
When asked what the key to developing a new binding affinity model is, Corso explains:
“Models such as Alphafold and Boltz are being developed. [a] They have a strong understanding of physical interactions without the need for simulation. Boltz-2 can exploit this understanding by training additional modules trained with millions of training data points from published biochemical assays. This allows [for] Accurate binding affinity calculations without the need for extensive simulations. ”
How Labs implements Boltz-2
The technical requirements for running Boltz-2 remain the same as those for Boltz-1. Researchers need to access a computer with modern GPUs (local or via the cloud). A GPU's 40 gigabytes of video RAM (VRAM) is ideal, but 32 or 24 gigabytes is sufficient for most input sizes. Corso points out that collaboration with NVIDIA has enabled Boltz developers to reduce model runtime and memory footprint. As a result, Boltz-2 can predict the structure of a larger complex than Boltz-1, even with the same amount of VRAM.
Interested lab managers can find source code and detailed installation instructions in the Boltz Github repository.
For those interested in keeping up with Boltz's development and connecting with other users, the Boltz team has a public rack channel to join.
