Prior-Fitted Functional Flow: In-Context Generative Models for Pharmacokinetics

2026-04-19

AI Summary

This work addresses the challenge of parameter-free population synthesis and individual trajectory prediction in pharmacokinetics by proposing an in-context generative model that, for the first time, integrates functional flows with population-level priors. By learning a function vector field conditioned on sparse and irregularly sampled observations, the method enables zero-shot generation of virtual populations and personalized pharmacokinetic trajectory forecasting. The approach unifies generative flow models, Bayesian priors, and irregular time-series modeling to deliver well-calibrated uncertainty estimates. It achieves state-of-the-art predictive accuracy across multiple real-world datasets and introduces the first open-source literature corpus specifically curated for pharmacokinetic priors.

Technology Category

• Reasoning under Uncertainty: Relational Probabilistic Models
• Machine Learning: Deep Generative Models & Autoencoders
• Natural Language Processing: Generation

Application Category

• User Modeling, Personalization and Recommendation: Explainable and interpretable methods for personalization
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Social Networks and Social Media: Generative AI / large language models and their impact on social systems

Abstract

We introduce Prior-Fitted Functional Flows, a generative foundation model for pharmacokinetics that enables zero-shot population synthesis and individual forecasting without manual parameter tuning. We learn functional vector fields, explicitly conditioned on the sparse, irregular data of an entire study population. This enables the generation of coherent virtual cohorts as well as forecasting of partially observed patient trajectories with calibrated uncertainty. We construct a new open-access literature corpus to inform our priors, and demonstrate state-of-the-art predictive accuracy on extensive real-world datasets.

Problem

• pharmacokinetics
• generative models
• population synthesis
• individual forecasting
• sparse irregular data

Innovation

• Prior-Fitted Functional Flows
• generative foundation model
• pharmacokinetics
• zero-shot population synthesis
• functional vector fields