The collaboration brings together Simulations Plus’ validated scientific engines across physiologically-based pharmacokinetics (PBPK), pharmacokinetics/pharmacodynamics (PK/PD), and quantitative systems pharmacology (QSP) with NVIDIA AI infrastructure to accelerate and scale simulation cycles, parameter exploration, and virtual population studies. Together, these capabilities address two core constraints in model-informed drug development (MIDD): reducing manual, expertise-driven work and enabling large-scale exploration of model structures and parameters in parallel—shifting modeling from a sequential process to a more iterative, data-informed workflow operating at program-relevant timelines.

NVIDIA contributes advanced computing infrastructure and expertise in accelerated inference and GPU-native optimization to improve simulation performance and enable interactive, AI-assisted workflows. NVIDIA also brings access to its life sciences ecosystem, including the NVIDIA BioNeMo platform and a global network of pharmaceutical partners, supporting broader engagement and adoption.

“For three decades, Simulations Plus has helped pharmaceutical and biotechnology organizations apply modeling and simulation with confidence across drug development. Our collaboration with NVIDIA brings together validated science, accelerated computing, and AI capabilities in a way that expands what scientific teams are able to explore and accomplish,” said Shawn O’Connor, Chief Executive Officer of Simulations Plus.

The collaboration is initially focused on three areas:

Next-generation scientific engines

Simulations Plus has begun developing GPU-optimized simulation capabilities for QSP and PK/PD applications, reducing runtimes and enabling broader exploration of complex biological systems. This enables scientists to evaluate a broader range of hypotheses without pre-pruning models, increasing confidence in model selection and supporting more robust program decisions.

AI-assisted scientific workflows

Simulations Plus is applying AI-assisted approaches, informed by NVIDIA’s expertise in accelerated inference and agentic AI, to support model construction, parameter fitting, diagnostics, and refinement—reducing manual effort and accelerating iteration from question to analysis. By reducing manual, time-intensive steps, these workflows will allow scientists to focus more on scientific interpretation and decision-making, significantly accelerating iteration cycles within drug development programs.

Advancing quantitative systems pharmacology

Simulations Plus is prioritizing QSP workflows—one of the most computationally demanding areas in drug development—by applying GPU acceleration and AI-assisted methods to improve simulation efficiency and expand practical use in pharmaceutical R&D. Current testing shows up to a 75% reduction in time required for end-to-end QSP modeling, enabling faster iteration and expanding the practical use of QSP within program timelines.

As part of the collaboration, the companies plan to engage select pharmaceutical partners to evaluate these capabilities in real-world drug development workflows, with initial focus on high-complexity modeling use cases.

“Scientific teams are asking for faster iteration, greater scale, and better ways to work across increasingly complex modeling problems. By combining our validated scientific engines with AI-assisted workflows and accelerated computing, we are extending our platform into a more integrated modeling ecosystem—where workflows scale across domains like QSP while remaining grounded in reproducible, scientifically validated outputs,” said Erik Guffrey, Co-Chief Product and Technology Officer of Simulations Plus.

“Biopharma teams need platforms that can connect mechanistic modeling, AI, and high-performance simulation into workflows scientists can actually use. By bringing NVIDIA accelerated computing and AI infrastructure together with Simulations Plus’ deep expertise in model-informed drug development, we can help researchers run more complex models, explore larger design spaces, and move from insight to decision faster,” said Anthony Costa, Director of Digital Biology and Health, NVIDIA.

About Simulations Plus, Inc.

Simulations Plus is a global leader in model-informed and AI-accelerated drug development. We create value for our clients by accelerating the discovery, development, and commercialization of pharmaceuticals and other products through innovative science-based software and consulting solutions. For more information, visit www.simulations-plus.com.

Forward-Looking Statements

Except for historical information, the matters discussed in this press release are forward-looking statements that involve risks and uncertainties. Words like “believe,” “will,” “can,” “expect,” “anticipate” and similar expressions (or the negative of such terms, as well as other words or expressions referencing future events, conditions or circumstances) mean that these are our best estimates as of this writing, but there can be no assurances that expected or anticipated results or events will actually take place, so our actual future results could differ significantly from those statements. Factors that could cause or contribute to such differences include, but are not limited to: effectiveness of our new operational structure, our ability to maintain our competitive advantages, acceptance of new software and improved versions of our existing software by our customers, the general economics of the pharmaceutical industry, our ability to finance growth, our ability to continue to attract and retain highly qualified technical staff, market conditions, macroeconomic factors, and a sustainable market. Further information on our risk factors is contained in our quarterly, annual and current reports and filed with the U.S. Securities and Exchange Commission.

We have long relied on in vitro–in vivo correlations (IVIVC) as a framework to connect these domains. In principle, the concept is sound: if dissolution can be measured in vitro, it should be possible to predict how a drug behaves in vivo. In practice, however, this translation is often less reliable than we would like.

The question, then, is not whether dissolution matters. It clearly does. The question is whether our current methods capture the reality of how drug products behave once they enter the human body.

Where Traditional IVIVC Falls Short

Classical IVIVC approaches typically rely on either direct input of in vitro dissolution data or the application of scaling factors for the in vivo dissolution time and fraction absorbed. These  methods can work well under certain conditions, particularly when dissolution is not solubility limited and when the formulation excipients control drug release.

However, for many modern drug candidates—especially those with low solubility or complex formulation characteristics—these approaches frequently fail to provide consistent predictive performance. The reason is straightforward: in vitro dissolution experiments do not replicate in vivo conditions. Differences in pH, fluid composition, hydrodynamics, and volume can significantly alter dissolution behavior. Moreover, the presence of bile salt micelles and dynamic changes in the lumen physiology  introduce additional layers of complexity that are not captured in standard in vitro dissolution tests.

The Z-factor is a mechanistic model that accounts for formulation effects in vitro and in vivo. Dissolution predictions using Z-factor are sensitive to dose, volume and solubility differences between in vitro and in vivo conditions. Z-factor, however, often lacks the granularity required to describe complex dissolution mechanisms or media. They are not inherently designed to handle formulation comprising multiple polymorphs, different micelles or non-first-order dissolution profiles. As a result, we are often left with models that fit data, but do not fully explain it.

A Shift Toward Mechanistic Understanding

Over time, it has become clear that empirical correlations alone were not sufficient. What is needed is a more mechanistic representation of dissolution—one that reflects not just the experimental conditions, but the physical reality of the drug product itself.

This shift aligns with the broader evolution toward model-informed drug development. Modeling is no longer used solely to interpret data after the fact; it is increasingly expected to guide decisions, inform formulation strategies, and support regulatory submissions.

To do so effectively, our models must be grounded in how drug products actually behave.

Rethinking the Drug Product: Introducing P-PSD

One of the key insights in this area is that the drug product is not simply a collection of particles defined by the drug substance (DS). The manufacturing process—blending, granulation, compression and the choice of the excipients—fundamentally alters the physical characteristics of the drug substance within the drug product.

Particles may fracture or agglomerate. Surface properties may change. Wettability may improve or deteriorate. The spatial distribution of the drug within the formulation can influence how much surface area is effectively available for dissolution. DS sizing methods cannot predict or capture these changes and this is where the concept of a product particle size distribution (P-PSD) becomes invaluable.

Rather than describing the DS in isolation, the P-PSD represents the effective surface area of the drug within the finished product—the surface that is actually available for dissolution under in vitro and physiological conditions. Importantly, the P-PSD is not measured directly. It is inferred by fitting observed in vitro dissolution data of the drug product itself, using mechanistic equations that account for the underlying dissolution processes. In doing so, it captures the combined effects of formulation, manufacturing, and physicochemical interactions in a way that is directly relevant to modeling.

From Dissolution Data to Mechanistic Representation

In practical terms, the P-PSD approach uses observed dissolution profiles from specific drug product batches as input. These data are then used to estimate a particle size distribution that reproduces the observed release behavior.

This process is not simply curve fitting. The underlying dissolution model incorporates key mechanisms, including:

• Surface-area–dependent dissolution represented by multiple bins of spherical particles in which the drug pass is distributed
• Drug binding to micelles in the dissolution medium (the affinity of the drug to micelles is measured independently)
• Potential drug degradation once it is solubilized (degradation rate constant can be measured separately from stock solutions)
• The influence of micelle size and number (the critical micelle concentration and drug affinity to micelles and micelle size are measured separately)
• The presence of multiple polymorphs, each with distinct solubility and affinity to micelles
• The potential precipitation from one polymorph to another (directly resulting from the dissolution equation)

The result is a P-PSD that reflects the physical and chemical realities of the formulation in a given dissolution medium. Once fitted to a given dataset, the P-PSD can be validated against additional dissolution conditions and then used as an input for physiologically based biopharmaceutics models (PBBM). This is where its true value emerges.

Why P-PSD Improves Predictive Confidence

By explicitly representing the surface area available for dissolution for each polymorph of the formulation, P-PSD provides a more realistic foundation for predicting in vivo behavior.

Several advantages follow:

• Improved mechanistic fidelity
  The model reflects how the product actually dissolves, rather than relying on simplified assumptions.
• Batch-specific insight
  Variability between batches can be captured and evaluated, supporting more robust development strategies and enabling PBBM validation and waivers of clinical bioequivalence studies.
• Reduced over-parameterization
  The approach emphasizes parsimony, identifying the minimum level of complexity (bins of drug sizes) required to describe the system.
• Enhanced integration with PBBM
  Because P-PSD is grounded in physical reality, it translates more naturally into physiologically based models, where particles of drug are moved along the GI tract, with smaller highly soluble particles dissolving first, followed by larger lower solubility particles.

Ultimately, this leads to greater confidence—not only in the model itself, but in the decisions that depend on it.

Implications Across the Development Lifecycle

The impact of this approach extends across multiple stages of drug development.

In early development, P-PSD can support formulation screening by helping teams understand how processing decisions influence dissolution behavior. This comprises the choice of the right polymorph and size of the drug substance (through its impact on drug product dissolution).

Half way through development, it provides a more reliable basis for establishing in vitro–in vivo relationships, and allow PBBM to be developed. Because of its mechanistic nature, fitting can work both ways: in vivo PK profiles could be fitted with a P-PSD if in vivo dissolution is limiting absorption and in vitro dissolution of that drug product could be predicted, thereby increasing the likelihood and speed to develop a biopredictive dissolution method.

In later stages, it strengthens the scientific foundation of regulatory submissions by offering a mechanistic rationale for predicted in vivo performance, supporting formulation changes, waiving unnecessary human evaluation and helping with selected quality specifications of the drug product, such as the dissolution specification, polymorphic impurity level specification, or drug substance particle size specification.

Across all stages, the goal is the same: reduce uncertainty and improve the quality of decision-making.

From Correlation to Understanding

The evolution from empirical correlation to mechanistic understanding is not unique to dissolution modeling. It reflects a broader trend across pharmaceutical science.

As our tools become more sophisticated and our data more abundant, the expectation is no longer simply to integrate observations. It is to explain them—and to use that understanding to predict what will happen under new conditions.

In this context, approaches like P-PSD represent an important step forward.

They do not replace experimental data, nor do they eliminate uncertainty entirely. What they offer is a more faithful representation of the system we are trying to model.

And in drug development, fidelity matters.

Learn more about P-PSD

Purpose: To develop and verify a physiologically based pharmacokinetic (PBPK) modeling strategy for mithramycin (MTM) and its analog, MTMSA-Trp, with the aim of projecting first-in-human plasma pharmacokinetics and supporting the translational development of MTMSA-Trp for Ewing sarcoma treatment.

Methods: PBPK models were created in GastroPlus® using a middle-out approach, incorporating preclinical pharmacokinetic data from mice, rats, and cynomolgus monkeys. Human clearance was estimated through three methods: an additional clearance approach, allometric scaling, and single-species scaling from monkeys. The model was evaluated using clinical MTM plasma PK data and then employed to project human MTMSA-Trp plasma PK, with tissue predictions considered exploratory.

Results: The additional clearance approach provided the most accurate prediction of human MTM plasma PK. Across all clearance prediction methods, MTMSA-Trp was predicted to achieve 8- to 15-fold higher human plasma exposure than MTM at the same dose. Model-derived liver exposures were 2- to 4-fold higher, with a lower predicted liver partition for MTMSA-Trp; however, these tissue predictions remained sensitive to distribution assumptions. Parameter sensitivity analysis identified the blood-to-plasma ratio as the most influential parameter among those examined.

Conclusion: PBPK modeling supports the projection that MTMSA-Trp will achieve substantially higher plasma exposure than MTM in humans. This empirically developed workflow may inform translational efforts for the first-in-human development of MTMSA-Trp.

By Kumar Kulldeep Niloy, Jamie Horn, Nazmul Hasan Bhuiyan, Suhas S. Bhosale, Khaled A. Shaaban, Thomas E. Prisinzano, Jon S. Thorson, Jurgen Rohr, Markos Leggas

• Multiple AIDD runs were performed with varying parameters; results were combined and AIDD-generated compounds were identified in the Enamine REAL and WuXi GalaXi synthesis-on-demand libraries.
• Initial results were used to rebuild QSAR models and a second round of AIDD was performed and compounds identified in synthesis-on-demand libraries.
• In all, 69 molecules were synthesized and tested.

By Jeremy O Jones, Rafal A Bachorz, Joanna Pastwińska , Anna Sałkowska, Marcin Ratajewski

AACR Annual Meeting 2026, April 17th – 22nd, 2026, San Diego, CA

“Complex oral formulations such as amorphous solid dispersions present significant scientific and regulatory challenges due to their sensitivity to physiological and manufacturing variables,” said Dr. Viera Lukacova, Chief Scientific Officer of Simulations Plus. “Through this funded collaboration, we aim to integrate advanced in vitro systems with mechanistic modeling to improve prediction of in vivo performance, support regulatory decision-making, and enable more efficient development pathways for these high-impact therapies that deliver faster dissolution and more drug absorption.”

Advancing Mechanistic, Model-Informed Approaches for Complex Products

ASDs are among the most powerful yet complex oral drug delivery systems, with performance influenced by factors such as food intake, gastric pH, formulation composition, and manufacturing processes. Current regulatory approaches often require multiple clinical bioequivalence (BE) studies, which can be resource-intensive while still carrying uncertainty.

The collaboration evaluates whether advanced in vitro dissolution systems—particularly those incorporating dynamic gastrointestinal physiology—combined with mechanistic physiologically based biopharmaceutics modeling (PBBM), can reliably predict key in vivo outcomes, including food effects and the impact of elevated gastric pH conditions.

By establishing and validating these predictive capabilities, the collaboration aims to provide a scientific foundation for reducing reliance on certain clinical BE studies while maintaining the rigor and transparency required by regulators.

Integrating Experimental and Mechanistic Modeling Expertise

The collaboration brings together complementary capabilities across experimental science and computational modeling.

Lonza will lead experimental work, including in vitro dissolution testing under fasted, fed, and elevated gastric pH conditions using advanced systems such as Controlled Transfer Dissolution (CTD), as well as the characterization and, where needed, manufacturing of ASD formulation variants.

Simulations Plus will lead the development and validation of in vitro–in vivo extrapolation (IVIVE) frameworks using its DDDPlus® and GastroPlus® platforms, translating experimental data into predictions of in vivo pharmacokinetics and supporting virtual bioequivalence assessments. At the same time, it creates new opportunities to extend these capabilities into grounded AI-enabled workflow environments, where data, mechanistic models, and simulation outputs will be more directly connected. The Company will also contribute to interpretation within a regulatory context, ensuring alignment with evolving expectations for model-informed drug development (MIDD).

Francois Ricard, Head of R&D, Lonza Advanced Synthesis, said, “This collaboration reflects Lonza’s commitment to advancing more predictive, science-driven approaches as the leader in the field of bioavailability enhancement. By combining advanced in vitro experimentation with mechanistic modeling, and working closely with Simulations Plus and the FDA, we aim to strengthen the scientific foundation that underpins regulatory decision-making for complex oral drug products. Ultimately, this type of collaboration should help accelerate development for our customers requiring bioequivalence during clinical development.”

Alignment with Regulatory Priorities and Industry Needs

This work is supported in part through FDA funding and includes ongoing engagement with FDA scientists to directly align with regulatory priorities to advance MIDD, modernize bioequivalence assessment for complex products, and reduce unnecessary reliance on human studies. By combining regulatory collaboration with open, non-proprietary data and validated methods based on real-world, FDA-approved ASD products, the initiative is intended to inform future regulatory approaches and support broader adoption of science-based alternatives.

“The industry is moving toward a future where decisions are informed earlier, with greater confidence and scientific transparency,” added Lukacova. “Our role is to ensure those decisions are grounded in validated science—while enabling more efficient ways to connect data, models, and insight.”

More than 1,400 scientists from industry, academia, and regulatory agencies from over 65 countries participated in the week-long program, reflecting the high demand for expert-led training as modeling and simulation increasingly become standard for drug development strategy, regulatory engagement, and clinical execution.

“As model-informed approaches become central to how therapies are developed, it is vital to apply modeling and simulation consistently across teams,” said Jonathan Chauvin, Co-Chief Product and Technology Officer of Simulations Plus. “Through programs like Spring School, we are expanding access to these capabilities to support the next generation of scientists, while enabling broader adoption of model-informed workflows across the industry. These foundational skills are essential for scientists to evolve with the discipline and leverage validated scientific engines and AI-enabled ecosystems to support better decision-making across the drug development lifecycle.”

Held from March 23 to 27, 2026, the Spring School program offered two tracks: GastroPlus® Spring School: From PBPK Basics to Biopharmaceutics Applications, and MonolixSuite™ Spring School: High-Impact Pharmacometrics Case Studies. Both tracks included interactive lectures, hands-on exercises, and live Q&A sessions led by Simulations Plus experts.

“The continued scale and diversity of participation in this year’s program reflects how quickly model-informed approaches are becoming embedded across the global scientific community,” said Jennifer Johnson, Manager of Learning Services of Simulations Plus. “Our focus is on building programs that not only educate but also help scientists translate these methods into real-world workflows and collaborative environments.”

Simulations Plus has a long-standing commitment to education. In addition to its Spring School, the Company has previously offered Summer and Autumn Schools focused on PK/PD modeling. The most recent Autumn School was the first program to offer a second track focused on PBPK modeling. In addition to these programs, Simulations Plus is widely known for its University+ program, which provides free academic access to modeling and simulation software for thousands of students and educators worldwide. Together, these initiatives form a cornerstone of the Company’s global educational outreach—helping expand the adoption of model-informed approaches and strengthening the pipeline of scientists equipped to apply these methods across the drug development lifecycle.

This study addresses the critical need for improved physicochemical descriptors and predictive models tailored to beyond Rule of Five (bRo5) compounds, including PROTACs and cyclic peptides. By integrating experimental parameters such as EPSA1, ChromLogD2, and ChameLogK3 with computationally derived features of molecular chameleonicity, we aim to establish new structure–property relationships that better capture the dynamic polarity and conformational adaptability of these complex molecules. The resulting models are designed to enhance the prediction of key in vitro ADME and in vivo pharmacokinetic (PK) endpoints, ultimately providing a more reliable framework for the design and optimization of developable bRo5 therapeutics. Case studies demonstrate the contribution of these new descriptors to model building as well as improvements in predicting key in vivo endpoints for this challenging chemical space.

By Jeremy O. Jones, Rafał A. Bachorz, Vladimir Chupakhin, Michael Lawless, David Miller, and Robert Fraczkiewicz

CHI Drug Discovery Chemistry, April 13th – 16th, 2026, San Diego, CA