Correlating in vitro performance with physico-chemical characteristics of nanofibrous scaffolds for skin tissue engineering using supervised machine learning algorithms

Abstract

The engineering of polymeric scaffolds for tissue regeneration has known a phenomenal growth during the past decades as materials scientists seek to understand cell biology and cell–material behaviour. Statistical methods are being applied to physico-chemical properties of polymeric scaffolds for tissue engineering (TE) to guide through the complexity of experimental conditions. We have attempted using experimental in vitro data and physico-chemical data of electrospun polymeric scaffolds, tested for skin TE, to model scaffold performance using machine learning (ML) approach. Fibre diameter, pore diameter, water contact angle and Young's modulus were used to find a correlation with 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT) assay of L929 fibroblasts cells on the scaffolds after 7 days. Six supervised learning algorithms were trained on the data using Seaborn/Scikit-learn Python libraries. After hyperparameter tuning, random forest regression yielded the highest accuracy of 62.74%. The predictive model was also correlated with in vivo data. This is a first preliminary study on ML methods for the prediction of cell–material interactions on nanofibrous scaffolds.

Keywords

• supervised learning algorithms
• skin tissue engineering
• polymeric scaffold performance
• cell–material interaction
• predictive model