Data-Driven Inverse Design of Electrospun Biomaterials with Controlled Fibre-Diameter Distributions for Tissue Regeneration

Electrospun nanofibrous scaffolds are widely used in tissue engineering because their fibrous architectures closely resemble the extracellular matrix (ECM), providing structural and biochemical cues that regulate cell adhesion, proliferation, and tissue regeneration. However, designing electrospun biomaterials with targeted structural features remains largely empirical. In particular, biological performance is strongly influenced by the full fibre-diameter distribution, which governs pore size, mechanical behaviour, and cell infiltration, yet most optimisation strategies rely only on mean fibre diameter.

This work presents a data-driven framework for the inverse design of electrospun biomaterials with tailored fibre-diameter distributions relevant to regenerative applications. Using a curated meta-dataset containing more than 68,000 fibre-diameter measurements extracted from 1,778 datasets across 16 polymers commonly used in biomaterials, we develop predictive models capable of reconstructing complete fibre-diameter distributions from experimentally controllable electrospinning parameters, including polymer composition, solvent system, and electrohydrodynamic conditions. Model performance is rigorously evaluated computationally and empirically to ensure robust generalisation across different experimental sources.

To enable scaffold design, the predictive models are embedded within a chemically constrained and Inverse Monte Carlo framework that generates feasible polymer–solvent–process combinations capable of achieving user-defined fibre-diameter distributions while respecting experimentally validated compatibility constraints. This inverse design approach allows researchers to computationally explore process conditions that produce biomimetic architectures, including bimodal fibre distributions known to enhance cell infiltration and tissue integration.

By integrating interpretable machine learning, chemically informed constraints, and distribution-aware modelling, this framework provides a new pathway for the rational design of electrospun biomaterials for tissue regeneration. The approach reduces trial-and-error experimentation, accelerates scaffold optimisation, and supports the development of next-generation electrospun constructs with tunable structural properties tailored to specific regenerative medicine applications.