SandboxAQ Launches Virtual Screening Solution for GPCR Drug Discovery, Accelerated by NVIDIA BioNeMo

PALO ALTO, Calif., June 23, 2026 /PRNewswire/ -- SandboxAQ today announced a new virtual screening solution that helps drug developers identify promising candidates faster and at lower cost for G protein-coupled receptors (GPCRs), one of the most valuable yet hardest-to-drug families of targets. Accelerated by NVIDIA BioNeMo Agent Toolkit, the GPCR Virtual Screening Solution predicts both whether a molecule binds a receptor and whether it is likely to activate the receptor or block its activation.

GPCRs are the targets of roughly a third of all approved drugs, including the GLP-1 medicines for diabetes and obesity, yet they remain notoriously difficult to drug. Because these receptors can shift between "on" and "off" states, a successful medicine must not only bind, but bind in a way that stabilizes the right state and produces the desired biological effect.

That is where the GPCR Virtual Screening Solution can help. It helps researchers predict what a GPCR ligand, whether a small molecule or peptide, will do – not just whether it binds. By forecasting how tightly a molecule binds to different receptor conformations and modeling its preference for the receptor's active versus inactive state, the Virtual Screening Solution infers functional activity and predicts whether a candidate acts as an agonist, antagonist, or inverse agonist. This physics-based approach connects static structural data to actionable pharmacology, helping teams reduce screening costs, shorten medicinal chemistry cycles, and reach candidate decisions with greater confidence before committing to extensive wet-lab testing.

"For decades, the field has largely been able to ask one question computationally: will this molecule bind?" said Andrea Bortolato, VP of Drug Discovery at SandboxAQ. "Another important question for a drug discovery program is what the molecule does once it binds. By modeling the physics of how a compound shifts a receptor between its active and inactive states, we can predict whether it activates the target or blocks its activation, and we can do it before molecules are made in the lab."

A single drug program can require synthesizing and testing thousands of compounds, each one costing time and money. Predicting how a molecule will behave before it is made lets teams narrow the field before they enter the lab, cutting wasted effort, lowering cost, and speeding the decisions that move a program forward.

Longer term, the same approach could enable discovery of biased ligands, allosteric modulators, and therapies against previously intractable or orphan GPCR targets. Ultimately, the value proposition is faster, more efficient drug discovery with higher confidence that compounds will produce the intended therapeutic effect.

"At SandboxAQ we see significant potential in the BioNeMo Agent Toolkit. We look forward to exploring the agent toolkit for accelerating our GPCR Virtual Screening workflows as well as additional projects," said Bill Fitzgerald, Vice President of Growth and Ecosystems at SandboxAQ.

SandboxAQ's Virtual Screening Solution Will Predict Mechanism of Action (agonist versus antagonist/inverse agonist)
In a retrospective mechanism-of-action benchmark, SandboxAQ's Virtual Screening Solution correctly identified every antagonist in the test set, separating these receptor-blocking molecules from activators by modeling how each compound shifts the receptor's energy between its active and inactive states.

It does so by following three steps:

• The first step is to generate biologically relevant GPCR structures. Many GPCR programs are limited by incomplete structural information. Even when receptor structures exist, the biologically relevant active and inactive conformations are often unavailable. SandboxAQ uses its large quantitative models (LQMs) to generate high-quality structural models of GPCR targets, including both active and inactive receptor states. The platform leverages multiple state-of-the-art protein structure prediction approaches and can incorporate additional biological context, such as G protein interactions, when necessary. The result is a structural foundation that enables downstream analysis of ligand (small molecule or peptide) binding and receptor activation.
• "An important part of GPCR discovery is getting the receptor structure right at the start," said Andrea Bortolato, VP of Drug Discovery at SandboxAQ. "Our workflow uses cofolding to model the active and inactive receptor states that determine downstream signaling behavior. Because we build on OpenFold3, we also benefit from advances such as cuEquivariance, and we are now testing TensorRT acceleration for Pairformer inference through our agent skills to make this step faster and more scalable."
• The second step is to rapidly identify likely binders (molecules that latch onto the target). To do so, machine learning rapidly screens the full compound library and flags likely binders, resulting in high accuracy and advancing only the most promising fraction. In the initial ML binder-screening step for a complex GPCR target, the method reached 79% accuracy and 83% specificity, meaning it not only made correct predictions in most cases, but was particularly effective at filtering out unlikely candidates before costly follow-up work.
• The third step is to predict the mechanism of action (agonist versus antagonist/inverse agonist). To do so, SandboxAQ applies rigorous physics-based modeling to that smaller set, sharpening the predictions and determining whether each leading compound stabilizes the active or inactive conformations of the target.

"What excites us most is where this leads," Bortolato added. "We're focused on extending the framework across more GPCR classes, and over time toward some of the hardest open problems in the field, including orphan receptors that have no known partner molecule and have so far resisted successful drug discovery. That's the direction we believe will reshape how these targets are pursued."

In the near term, SandboxAQ is focused on demonstrating the broad applicability of its agonist-versus-antagonist prediction framework across GPCR families. Building on that foundation, the company plans to expand the platform toward higher-order GPCR discovery challenges, including receptor deorphanization, identification of novel therapeutic targets, and systematic discovery of allosteric binding sites—areas where predictive, structure-based approaches could significantly reduce reliance on large-scale empirical screening and unlock new opportunities for therapeutic innovation.

SandboxAQ and NVIDIA Collaboration
NVIDIA's BioNeMo Agent Toolkit directly supports SandboxAQ's GPCR Virtual Screening solution, providing the computing foundation for its large-scale modeling and screening workloads.

"In addition, we are excited to explore if BioNeMo agents can help SandboxAQ accelerate an agentic GPCR screening capability," said Bill Fitzgerald, Vice President of Growth and Ecosystems, at SandboxAQ.

About SandboxAQ
SandboxAQ is a B2B company delivering solutions at the intersection of AI and quantum techniques. The company's Large Quantitative Models (LQMs) deliver critical advances in life sciences, financial services, navigation, and other sectors. SandboxAQ is an independent, growth-backed company funded by leading investors and strategic partners including funds and accounts advised by T. Rowe Price Associates, Inc., Google, Alger, IQT, US Innovative Technology Fund, S32, Paladin Capital, BNP Paribas, Eric Schmidt, Breyer Capital, Ray Dalio, Marc Benioff, Thomas Tull, and others. For more information, visit www.sandboxaq.com.

SOURCE SandboxAQ