Microfluidic production of silk fibroin nanoparticles: Process optimization and modeling

Silk fibroin is an attractive material for drug delivery due to its biocompatibility, stability, and ability to form nanoparticles through solvent-induced self-assembly. Although microfluidic nanoprecipitation offers superior control and reproducibility compared to batch methods, its rational optimization remains challenging because the tight coupling between flow dynamics, solvent exchange, and fibroin self-assembly hinders nanoparticle formation. Here, we present an integrated experimental-computational framework that combines Design of Experiments, machine-learning prediction, and computational fluid dynamics (CFD) to investigate how operating conditions influence the formation of silk fibroin nanoparticles (SFNs) within a specific microfluidic geometry. The combined effects of fibroin:acetone ratio (1:3–1:5) and total flow rate (8–140 mL/min) were systematically investigated, with particle size and distribution quantified by Nanoparticle Tracking Analysis. Linear, power-law, and Random Forest models captured non-linear interactions between operating parameters, identifying intermediate to high flow rates and higher acetone fractions as optimal for producing small, homogeneous nanoparticles. CFD simulations mechanistically rationalize these trends by quantifying mixing efficiency, shear rate, residence time, and solvent-rich regions that trigger fibroin nanoprecipitation. CFD-derived descriptors, including Volume of Change and process efficiency, delineated parameter regions associated with controlled desolvation and uniform nanoparticle formation. By integrating experimental data, data-driven modeling, and CFD-based mechanistic analysis, this study demonstrates how these complementary approaches can elucidate structure-process relationships in microfluidic SFN production. While the insights are system-specific, the proposed workflow provides a transparent and mechanistically grounded approach for interrogating complex nanoprecipitation processes and informing hypothesis-driven process refinement in related microfluidic systems.