AI Agents in the Lab: Simulations Plus and NVIDIA Redefine Drug Discovery

RESEARCH TRIANGLE PARK, NC – June 23, 2026 – In a move set to accelerate the fusion of artificial intelligence and pharmaceutical research, Simulations Plus today announced it is embedding NVIDIA’s new BioNeMo Agent Toolkit into its Composer software platform. The collaboration aims to create sophisticated AI “agents” capable of autonomously advancing drug development, marking a significant leap from predictive AI to proactive, agentic systems that can reason, plan, and execute complex scientific workflows.

The announcement, which coincided with NVIDIA’s unveiling of the toolkit at the BIO International Convention, expands a partnership first established in May 2026. It signals a strategic push to equip AI with the tools to not only process data but to actively participate in the scientific method, potentially slashing the time and cost of bringing new medicines to market.

The Dawn of the AI Research Assistant

At the heart of the initiative is the concept of “agentic AI.” Unlike previous generations of AI that required constant human guidance to perform tasks, these new agents are designed for greater autonomy. They can independently plan and execute multi-step research projects, from sifting through scientific literature to running complex simulations, effectively acting as tireless digital research assistants.

Simulations Plus is building this agentic layer for Composer, its AI-native platform for model-informed drug development, using the NVIDIA BioNeMo Agent Toolkit. This toolkit functions as a specialized library of skills for AI, giving general-purpose large language models the domain-specific knowledge needed to tackle tasks in molecular biology, genomics, and chemistry. It allows an AI to understand how to use digital tools, interpret scientific results, and even troubleshoot its own experiments.

By integrating these capabilities, Composer will host agents that coordinate intricate scientific workflows. “AI can help scientists explore more possibilities, but in drug development those insights must remain reproducible, explainable, and grounded in validated science,” said Erik Guffrey, Co-Chief Product and Technology Officer of Simulations Plus. “By combining the NVIDIA BioNeMo Agent Toolkit with our scientific engines, we’re building agents that seek to accelerate scientific work while preserving the rigor required for critical development decisions.”

Building Trust in the Digital Lab

One of the most significant hurdles for AI in high-stakes fields like medicine is the “black box” problem—when an AI provides an answer without a clear, verifiable path to its conclusion. This collaboration directly confronts this challenge through two key areas of focus.

The first is grounding the AI’s reasoning in trusted evidence. Composer agents will leverage NVIDIA Nemotron™ Parse, a sophisticated vision-language model designed to deconstruct and understand complex documents like scientific papers. Going beyond simple text extraction, Nemotron Parse identifies the structure of a document—distinguishing titles, tables, figures, and footnotes—to extract information with its original context intact. Crucially, it maintains provenance, creating a digital breadcrumb trail that allows a researcher to trace any AI-generated recommendation directly back to the source data or sentence in the original literature. This traceability is paramount for ensuring that AI-driven insights are not only innovative but also scientifically valid and defensible.

“Our goal is to help researchers move from question to insight more efficiently while keeping validated science at the center of decision-making,” stated Shawn O’Connor, Chief Executive Officer of Simulations Plus. This focus on explainability aims to build the trust necessary for scientists and regulatory bodies to rely on AI for critical R&D decisions.

Supercharging Complex Biological Simulations

The second pillar of the collaboration focuses on raw computational power, specifically for accelerating Quantitative Systems Pharmacology (QSP). QSP models are incredibly complex mathematical representations of how a drug interacts with biological systems, but their computational intensity has historically been a bottleneck. To address this, the companies are advancing nvQSP, a joint initiative to create CUDA-optimized solvers for the ordinary differential equations (ODEs) that form the backbone of QSP.

By running these simulations on NVIDIA’s accelerated computing infrastructure, the initiative drastically reduces the time required for these calculations. This acceleration enables scientists to explore a much wider range of possibilities—testing more drug candidates, evaluating more dosing regimens, and modeling more patient variations—than was previously feasible. This allows for a more thorough and rapid evaluation of a drug's potential efficacy and safety, helping researchers iterate faster and with greater confidence in their findings.

These accelerated QSP capabilities are being woven directly into the Composer platform, allowing the AI agents to initiate, manage, and interpret these complex simulations as part of their autonomous workflows. This synergy ensures that the agents' high-level reasoning is supported by rigorous, validated computational science at a massive scale.

A Strategic Alliance Reshaping the Landscape

The partnership between Simulations Plus, a leader in pharmaceutical modeling software, and NVIDIA, the dominant force in AI and accelerated computing, exemplifies a crucial trend in the biotech industry. As the complexity of biological data and AI models grows, strategic alliances that combine deep domain expertise with cutting-edge technology infrastructure are becoming essential for innovation.

In a competitive field populated by AI-native drug discovery firms like Insilico Medicine and Recursion Pharmaceuticals, as well as established software providers like Schrödinger, this collaboration solidifies Simulations Plus's position. It leverages its long-standing expertise in validated computational engines while integrating a powerful, best-in-class AI framework. This strategy avoids the need to build foundational AI models from scratch, instead focusing on the application and integration of these powerful tools to solve specific, high-value problems in drug development.

The collaboration represents a calculated move to transform drug development from a series of discrete, human-driven steps into a continuous, AI-accelerated cycle of discovery.