Becoming a Better Scientist with AI: How Tools Like GastroPlus X.2 and GastroPlusGPT Support Modelers

In physiologically based pharmacokinetic and biopharmaceutics modeling (PBPK/PBBM), we pride ourselves on precision, scientific rigor, and mechanistic exploration of complex biological systems. However, like in many scientific disciplines, we face a common set of practical challenges: extracting data from diverse sources, configuring complex models, troubleshooting errors, and translating results into decisions.

This is where artificial intelligence (AI) has begun to make a real, tangible difference: not as a replacement for scientific thinking, but as a companion to help facilitate it.

In this post, I will discuss how AI-powered tools-specifically GastroPlusGPT, a suite of AI-powered web apps released alongside GastroPlus 10.2 (GPX.2)-can help modelers work more efficiently, focus on deeper analysis, and ultimately be better scientists by working smarter.

Domain-Specific AI for Modeling Support

Unlike general-purpose AI platforms, GastroPlusGPT has been developed specifically for PBPK/PBBM modeling applications. Built on curated scientific content, including GastroPlus documentation, tutorials, and decades of collective industry experience and established modeling practices, these AI tools assist in model building and deliver guidance that is specific to each modeling stage.

GastroPlusGPT supports a variety of functions, including:

• Analyzing images of plots or tables to extract data from various sources, preparing it for download or entry into GastroPlus

• Answering questions about the configuration of complex models

• Organizing and enhancing the advice from the embedded knowledge in GastroPlus with explanatory text and additional plots to help with troubleshooting and decisions

• Interacting with the chatbot to answer any questions about the recommendations from GastroPlus and generate clear step-by-step instructions for implementation

By functioning as an integrated knowledge resource, the tool enables modelers to find answers more efficiently without disrupting their workflow or diverting attention to external documentation.

Streamlined Data Extraction from Unstructured Sources

A persistent challenge in PBPK/PBBM modeling is the time-consuming task of locating and extracting quantitative data from regulatory reports, literature, and internal documents. Manual data mining remains one of the most labor-intensive aspects of model development.

The AI-powered document extraction capability in GastroPlusGPT addresses this issue by automatically identifying and retrieving relevant data elements from uploaded unstructured files. This includes extraction of physicochemical properties, ADME parameters, and other critical model inputs, which can then be reviewed and integrated into GastroPlus simulations.

While final scientific judgment and validation remain the responsibility of the modeler, the reduction in manual transcription and formatting effort significantly improves the throughput of model development activities.

Reliability and Transparency in AI-Assisted Workflows

The introduction of AI into regulated or decision-informing scientific environments understandably raises questions regarding accuracy, traceability, and robustness. To that end, the development of GastroPlusGPT has been guided by scientific oversight. Its outputs are trained on validated knowledge sources, and its recommendations are grounded in standard modeling methodologies and based on the pearls of wisdom that have been hard-coded into GastroPlus.

AI as a Complement to Scientific Expertise

It is essential to emphasize that the tools described here automate; they don't replace the critical scientific reasoning process. AI enables scientists to devote more time to high-value tasks by eliminating routine inefficiencies. When thoughtfully implemented, AI acts as a force multiplier, facilitating deeper exploration of complex scenarios, faster iteration of model configurations, and more consistent application of best practices.

AI-enabled tools such as those integrated into GPX.2 represent a meaningful step forward in the practice of model-informed drug development. By focusing on practical challenges-data accessibility, software navigation, and real-time support-these tools help modelers maintain scientific rigor while accelerating the path from data to decision.

As a PBPK/PBBM scientist, I've long believed that the value we bring lies not just in the models we build, but in the questions we ask and the insights we generate. AI tools like GastroPlusGPT and the enhancements in GastroPlus 10.2 aren't just conveniences-they're accelerators for scientific discovery.

The future of modeling isn't about replacing people with algorithms. It's about equipping people with the smartest tools so the combined force of artificial and human intelligence can get treatments to patients faster.

Schedule a Demo of GastroPlusGPT