Montai Therapeutics leverages NVIDIA NIM for multimodal AI drug discovery.

Montai Therapeutics, a leading company, is making significant progress in drug discovery by leveraging a multimodal AI platform developed in collaboration with NVIDIA. According to the NVIDIA Technology Blog, this innovative platform uses NVIDIA NIM microservices to address the complexities of computer-based drug discovery.

The role of multimodal data in drug discovery

New drug development aims to develop new treatments that effectively target diseases while minimizing side effects for patients. The use of multimodal data such as molecular structures, cellular images, sequences, and unstructured data can be invaluable in identifying new and safe drug candidates. However, creating multimodal AI models poses challenges, including sorting different data types and managing significant computational complexity. Ensuring that these models effectively use information from all data types without introducing bias is a major challenge.

Montai’s innovative approach

Montai Therapeutics is using the NVIDIA BioNeMo platform to overcome these challenges. At the core of Montai’s innovation is the integration and curation of the world’s largest Anthromolecule chemical library. Anthromolecules refers to a rigorously selected collection of bioactive molecules consumed by humans from foods, supplements, and herbal medicines. This diverse chemical source provides significantly greater chemical and structural diversity than existing synthetic combinatorial chemical libraries.

Anthromolecules and their derivatives have already proven to be a source of FDA-approved drugs for a variety of diseases, but have not yet been exploited for systematic drug development. The abundance of topological structures across this diverse chemistry provides a much broader range of vectors to combine complex biology with precision and selectivity, potentially unlocking small molecule pill-based solutions for targets that have historically eluded drug developers. .

Creating a multimodal AI platform

In a recent collaboration, Montai and the NVIDIA BioNeMo Solutions team developed a multimodal model aimed at virtually identifying potential small molecule drugs from Anthromolecule sources. Built on AWS EC2, the model has been trained on several large biological datasets. It integrates NVIDIA BioNeMo DiffDock NIM, a state-of-the-art generative model for blind molecular docking pose estimation. BioNeMo DiffDock NIM is part of NVIDIA NIM, a set of easy-to-use microservices designed to accelerate generative AI deployments across clouds, data centers, and workstations.

This collaboration has resulted in notable model architecture optimizations in the backbone of contrastive learning-based models. Initial results are promising, with the model showing better performance than existing machine learning methods for predicting molecular function. Multimodal models integrate information across four modalities:

1. chemical structure
2. Phenotypic cell data
3. gene expression data
4. Information about biological pathways

Combining these four modalities resulted in a model that outperformed single modality models, demonstrating the benefits of contrastive learning and foundational model paradigms in AI for the drug discovery space.

By integrating these different modalities, this model will help Montai Therapeutics more effectively identify promising lead compounds for drug development through the CONECTA platform. This innovative drug operating system will facilitate the predictable discovery of innovative small molecule drugs from a wide range of untapped human chemistries.

future direction

Current joint efforts are focused on integrating a fifth modality derived from DiffDock predictions: the “docking fingerprint.” NVIDIA BioNeMo’s role has been instrumental in scaling the inference process to enable more efficient computation. For example, DiffDock on the DUD-E dataset using 40 poses per ligand on eight NVIDIA A100 Tensor Core GPUs achieves a processing speed of 0.76 seconds per ligand.

These advances highlight the importance of efficient GPU utilization in drug screening and highlight the successful use of NVIDIA NIM and multimodal AI models. The collaboration between Montai and NVIDIA represents an important step forward in the pursuit of a more effective and efficient drug discovery process.

Learn more about NVIDIA BioNeMo and NVIDIA BioNeMo DiffDock NIM.

Image source: Shutterstock