Risk substance identification of asphalt VOCs integrating machine learning and network pharmacology

Abstract

Asphalt releases volatile organic compounds (VOCs) during paving processes, posing risks to workers and the environment. The complex composition of asphalt and the evolving of VOCs present challenges in accurately assessing their potential environmental and health impacts using traditional experimental approaches. This study aimed to develop a robust computational framework integrating machine learning and network pharmacology to predict the risks from the asphalt VOCs. The results show that the MACCS+XGBoost model achieved the highest predictive performance, with an accuracy of 0.85, balanced accuracy of 0.84, sensitivity of 0.83, specificity of 0.84, and F1-score of 0.84 in the external validation. The network pharmacology analysis revealed that the identified VOCs with reproductive toxicity potential may disrupt key processes such as spermatogenesis, ovarian function, and hormonal regulation, providing mechanistic insights into their potential impacts. This advancement supports a proactive approach to environmental protection and fosters the transition towards a more sustainable, low-carbon transportation.