• Neo-1 is the first model to unify de novo molecular generation and atomic-level structure prediction in a single model, by generating latent representations of whole molecules instead of predicting atomic coordinates
• Neo-1 is generalizable across structural and sequence inputs, including partial inputs, and allows for arbitrary constraints with real-world empirical data, significantly reducing the search space needed to identify small molecule drug candidates, and thus dramatically improving the empirical value of computational drug discovery
• Optimized together with NVIDIA, Neo-1 enables design for high-value mechanisms of action, including molecular glues and other proximity-based therapeutics, directly addressing one of pharma’s biggest discovery bottlenecks, and is already in use across three large pharma and biotech partnerships with VantAI

NEW YORK: #AI--VantAI today unveiled Neo-1, the world’s most advanced atomistic foundation model and the first to unify de novo molecular generation with multimodal structure prediction, at NVIDIA GTC 2025.

Neo-1 is designed to make the systematic re-wiring of protein interactions possible, enabling drug developers to precision-engineer novel therapeutics, including molecular glues, bifunctional degraders, and other proximity-based modalities. For an expanded technical preview, click here.

“Our vision is to make protein interactions programmable,” said Luca Naef, Co-Founder & CTO of VantAI who led the development of Neo-1. “This is only possible if we can precision-design the structure of novel biomolecules, including drug-like small molecules. Existing models allow us to decode the structures of molecules, but to make new medicines we need to be able to design them. Neo-1, for the first time, allows for this in a single unified model.”

Neo-1 represents a novel approach for the field. While prior all-atom structure foundation models merely predict atomic coordinates, Neo-1 generates latent representations of molecules which can then be decoded into their three-dimensional structure. This methodology enables Neo-1 not only to predict the structure of protein-small molecule complexes to generate entirely new molecules tailor made to a given pocket.

As a result, Neo-1 unlocks new opportunities in drug design, including for novel molecular glues and other emerging proximity modulation therapeutics—modalities in which VantAI is leading innovation.

Critical advances underlying Neo-1 include:

• Multimodal inputs — accepts any combination of (partial) sequence, (partial) structure, and experimental data, such as NeoLink-derived restraints or almost any real-world empirical constraint
• Unified de novo molecule generation and structure prediction — enables atomic-resolution design and reconstruction of protein interfaces, critical for challenging mechanisms of action (MOAs) like molecular glues
• Massive-scale training — among the largest diffusion-based models in biology, trained on structural and synthetic datasets using hundreds of NVIDIA H100 GPUs
• Custom datasets and tools — incorporates PINDER & PLINDER (co-developed with NVIDIA) and synthetic data created using GPU-mmseqs2, a tool developed by NVIDIA and SNU, now planning on deploying it via NVIDIA’s recently announced MSA-Search NIM

In particular, Neo-1 has been trained using VantAI’s proprietary NeoLink dataset. This access to a vast new type of data and broad programmability has yielded a model that is highly performative, enabling precise molecular design and reconstruction of protein interface structures at atomic resolution, joint capabilities that are critical for unlocking de novo design of challenging MOAs such as molecular glues. VantAI has leveraged this to generate active small molecules for previously undruggable targets in its pipeline in a time frame of weeks, instead of what has traditionally taken years.

Michael Bronstein, Chief Scientist at VantAI and DeepMind Professor of AI at the University of Oxford noted, “Neo-1 unlocks the potential of VantAI’s high-throughput ‘black-box’ data platform, NeoLink. VantAI’s unique approach to data and AI will allow us to illuminate the geometric vocabulary of protein interaction at the full proteome scale, a step necessary to design next-generation therapeutics currently beyond reach. Being both one of the leading frontier AI labs in the bio space and having such a differentiated data source makes VantAI truly unique.”

“Neo-1 is a breakthrough model that enables all-atom de novo design across biomolecular space in addition to remarkable multimodal structure prediction capabilities,” said Anthony Costa, Director of Digital Biology at NVIDIA. “Acceleration of Neo-1 on the NVIDIA platform is crucial to accelerate scientific advances in the hands of scientists using the model.”

VantAI is currently deploying Neo-1 to select commercial and academic partners, including Janssen (J&J), Blueprint Medicines, and Bristol Myers Squibb, and looks forward to leveraging Neo-1 to accelerate key milestones across these successful discovery collaborations as well as for VantAI’s own internal pipeline.

About VantAI

VantAI is pioneering a transformative approach to drug discovery by illuminating the dynamic networks of protein interactions that drive biological function. Leveraging our groundbreaking NeoLink structural proteomics platform and next generation foundation models, we generate unprecedented interactomics data at scale, revealing how molecules connect, interact, and shape cellular behavior — in 3D. Alongside a robust and mechanistically differentiated internal pipeline, VantAI collaborates extensively with external partners to accelerate the delivery of innovative therapies. Our cross-disciplinary team of scientists and engineers is uniquely positioned to unlock novel therapeutic strategies that induce, modulate, or block protein interfaces—making previously undruggable targets accessible and providing powerful new ways to target diseases by rewiring cellular circuitry.

For more information, visit www.vant.ai.

Fonte: Business Wire