Machine Learning Integrated Workflow for Predicting Schwann Cell Viability on Conductive MXene Biointerfaces

Area of Science:

• Biomaterials Science
• Neuroscience
• Machine Learning

Background:

• Peripheral nervous system (PNS) injuries necessitate Schwann cell involvement for neuronal regeneration.
• Low-frequency electrical stimulation promotes neuron-Schwann cell co-growth in injured PNS.
• The precise relationship between electrical stimulation and Schwann cell viability remains unclear.

Purpose of the Study:

• To develop a machine learning (ML)-integrated workflow to assess Schwann cell viability.
• To investigate the effects of fabrication parameters and electrical stimulation on cell viability using conductive hydrogel biointerfaces.
• To create a predictive model for Schwann cell viability based on experimental parameters.

Main Methods:

• Fabrication of a hydrogel array with varying MXene and peptide loadings as conductive biointerfaces.
• Incubation of Schwann cells and application of varied electrical stimulation (voltage, frequency).
• Development of an artificial neural network model trained with experimental data and augmented with 1000-fold virtual data points.

Main Results:

• A high-accuracy prediction model for Schwann cell viability was achieved (testing mean absolute error ≤11%).
• The model accurately predicts cell viability based on fabrication and stimulation parameters.
• SHapley Additive exPlanations provided data-scientific insights validated by cellular observations.

Conclusions:

• A hybrid approach combining conductive biointerfaces, ML, and data analysis offers a novel platform for preclinical cellular-level prediction.
• This workflow enables accurate prediction of Schwann cell viability, crucial for advancing PNS regeneration strategies.
• The study provides a foundation for optimizing biointerface design and electrical stimulation protocols for nerve repair.