Optimization of printability of bioinks with multi-response optimization (MRO) and artificial neural networks (ANN)

In bioprinting, printing resolution and structural stability depend closely on the bioinks’ rheological properties such as zero shear viscosity, storage modulus, thixotropic recovery, viscoelasticity, and gelation point. Thus, understanding the material-rheology-printability relationships is crucial for multi-material bioinks. This study adopted a design of experiment (DoE) with response surface methodology using a central composite design to systematically investigate the rheological and printability parameters of bio-inks formed through combinations of sodium alginate, gelatin, and a nano-clay reinforcing agent (laponite) for enhanced storage modulus and cellular attachment. The material composition for the optimal printability was determined by the multi-response optimization method. Furthermore, this study incorporated machine learning techniques to generalize the effects of various rheological properties on printability and extrusion pressure. Multi-objective optimization was employed to statistically optimize solution properties based on the two opposing parameters: printed structure conformity and minimum extrusion pressure. The optimized bioinks demonstrated high-fidelity printing performance: less than 5% deformation from the computer-aided-design (CAD) models at low extrusion pressures below 30 Kpa for maintaining good cell viability. Resampling data from the DoE-fitted model equations facilitated the generation of extensive datasets for training artificial neural network (ANN) models. This process resulted in a robust machine learning model capable of accurately predicting bioink printability with a maximum 6.3% mean absolute error (MAE) solely based on the rheological properties. In summary, the DoE-based data sampling, MRO optimization, and ML modeling approach enabled the development of a robust bioink formulation method applicable to creating bioinks with extreme properties. The study underscores the crucial role of data-driven modelling and optimization approaches in extrusion-based bioprinting for tissue engineering applications.