Transfer learning for micro- and nanoplastic toxicity

A study led by postdoctoral researcher Kunpeng Chen, “Transfer Learning Enables Robust Prediction of Cellular Toxicity from Environmental Micro- and Nanoplastics,” was published in the Journal of Hazardous Materials.

The study developed a transfer learning-based QSAR model to address limited toxicity data for micro- and nanoplastics. The approach used pretraining on larger nanoparticle datasets and fine-tuning with smaller micro- and nanoplastic datasets, incorporating experimental and quantum chemistry features.

The study reported improved cytotoxicity prediction relative to conventional machine learning models, identified surface functionalization and dose as important predictors, and suggested that nanoplastics may induce cytotoxicity more rapidly than microplastics.