Development and predictive modeling of PLLA/gelatin nanofibers with functionalized CNTs for bone tissue engineering

Polymeric scaffolds are crucial for addressing bone defects. In this study, electrospun scaffolds composed of poly( L -lactic acid) (PLLA) and gelatin (Gel) in a 50:50 ratio were fabricated with varying concentrations of carbon nanotubes (CNTs) (1, 3, 5, and 7 wt%) and phosphate-modified CNTs (PCNTs). Electrospinning parameters were optimized to enhance fiber morphology. Nanofibers generated from a 12 % w/v PLLA/Gel solution containing 3 wt% CNT or PCNT exhibited uniform morphologies, with average diameters of 301 ± 93 nm and 296 ± 81 nm for PLLA/Gel/CNT and PLLA/Gel/PCNT, respectively. The scaffolds' physical, mechanical, and biological properties were systematically evaluated. Statistical analyses were employed to characterize physicochemical features, and machine learning (ML) models were developed to predict scaffold performance in bone tissue engineering. Six supervised ML algorithms were trained using the Seaborn and Scikit-learn Python libraries. Model evaluation using 5-fold cross-validation revealed that the Support Vector Regressor (SVR) achieved the best overall balance between error reduction and model fit, while Random Forest showed the lowest prediction error. Results revealed that CNT incorporation improved piezoelectric behavior, enhanced mechanical strength, and accelerated scaffold biodegradability. Additionally, biocompatibility assessments confirmed the non-toxicity of CNT-containing scaffolds. This work not only elucidates the physicochemical characteristics of electrospun composites but also introduces predictive ML methodologies for modeling cell-material interactions, offering a robust framework for advancing bone tissue engineering applications.