Leveraging GenAI for the automation of Model Based Meta Analysis (MBMA)
Emily Nieves, Co-founder and CEO, Delineate Inc., Cambridge, MA
Emily Nieves holds a Bachelor of Science in Biological Engineering from the University of Georgia, where she conducted research in Dr. Melissa Hallow's Quantitative Systems Pharmacology lab. Her professional experience includes the development of Quantitative Systems Pharmacology and machine learning models across diverse therapeutic areas, including the cardiorenal system, metabolism, and cell-based therapies, with contributions at Pfizer and AstraZeneca.
As a former PhD candidate in Biological Engineering at MIT, Emily specialized in hybrid AI and scientific machine learning algorithms. She currently serves as Co-Founder and CEO of Delineate, a seed-stage company that collaborates with pharmaceutical organizations to leverage large language models and artificial intelligence technologies in advancing Model Informed Drug Development workflows.
This presentation will demonstrate how Delineate's software applies generative AI to streamline Model Informed Drug Development (MIDD) workflows, with particular emphasis on Model Based Meta-Analyses (MBMA). The platform automates four critical phases of MBMA preparation: systematic literature review, data extraction from publications, creation of harmonized, analysis-ready datasets, and model building.
The software automatically digitizes data from plots, graphs, and tables within published literature, extracting dosing regimens, observation events, and covariates with minimal manual intervention. Extracted data are transformed into flexible formats compatible with NONMEM and similar pharmacometric software platforms, significantly reducing the time and resources required for database preparation. AI agents then leverage the standardized dataset, context from the publications, and MBMA best practices to assist the user in the model building and testing phase.
Beyond demonstrating efficiency gains in MIDD applications, this talk addresses key challenges broadly relevant to generative AI implementation in scientific workflows. Topics include strategies for efficient quality control of AI-extracted data and methods for standardizing heterogeneous, unstructured data sources into analysis-ready formats. These approaches are applicable to MBMA and other data-intensive pharmacometric analyses, offering a scalable solution to one of the field's most time-consuming bottlenecks.