Abstract

RECENT ADVANCEMENT IN USE OF COMPUTER IN PHARMACEUTICAL FORMULATION DEVELOPMENT

Shreya D. Palase*, Yogesh B. Raut, Sanjay K. Bais

ABSTRACT

Pharmaceutical formulation development is undergoing rapid transformation through computational innovations that streamline design, minimize experimental workload, and strengthen product reliability. Progress in this field is being driven by both data-centric and mechanistic strategies. Artificial intelligence method like data driven learning and Bayesian optimization act as increasingly applied examine historical datasets, high throughput outputs, enabling prediction of ideal excipient compositions, processing conditions, and critical quality attributes, while active learning approaches strategically reduce the number of required experiments. Alongside these, mechanistic and multiscale simulations— encompassing physiologically based pharmacokinetics (PBPK) and PK/Pattern, dissolution–permeation integration, computational fluid dynamics for process operations, and discrete element methods at the particle scale—offer in silica insights into bioavailability, stability, and scale-up behavior. The emergence of digital twins, combining sensor data with predictive models, is advancing real-time monitoring and paving the way for real-time release testing. For advanced dosage forms such as lipid nanoparticles, liposomes, amorphous dispersions, and long-acting injectable, molecular modeling and predictive stability tools guide excipient selection and manufacturing feasibility. Furthermore, integration with automated systems, electronic laboratory notebooks, and FAIR compliant data infrastructures enhances traceability and iterative model refinement. Broader implementation is being supported by model-informed drug development frameworks, though challenges remain—including limited and variable datasets, difficulties in scaling models across processes and sites, uncertainty management, and regulatory requirements for validation and lifecycle oversight. Looking forward, breakthroughs such as foundation models for formulation science, physics-informed neural networks, federated learning for collaborative progress, and stronger integration between PAT and AI systems are expected to deliver faster, more sustainable, and more dependable formulation development across the pharmaceutical pipeline. 

Keywords: Formulation informatics; machine learning; Bayesian optimization; PBPK; QbD/DoE; multiscale modeling; digital twins; PAT; real-time release testing; excipient screening; stability prediction; computational fluid dynamics.