Sensitivity Evaluation of Enveloped and Non-enveloped Viruses to Ethanol Using Machine Learning: A Systematic Review

Abstract

Viral diseases are a severe public health issue worldwide. During the coronavirus pandemic, the use of alcohol-based sanitizers was recommended by WHO. Enveloped viruses are sensitive to ethanol, whereas non-enveloped viruses are considerably less sensitive. However, no quantitative analysis has been conducted to determine virus ethanol sensitivity and the important variables influencing the inactivation of viruses to ethanol. This study aimed to determine viruses’ sensitivity to ethanol and the most important variables influencing the inactivation of viruses exposed to ethanol based on machine learning. We examined 37 peer-reviewed articles through a systematic search. Quantitative analysis was employed using a decision tree and random forest algorithms. Based on the decision tree, enveloped viruses required around ≥ 35% ethanol with an average contact time of at least 1 min, which reduced the average viral load by 4 log₁₀. In non-enveloped viruses with and without organic matter, ≥ 77.50% and ≥ 65% ethanol with an extended contact time of ≥ 2 min were required for a 4 log₁₀ viral reduction, respectively. Important variables were assessed using a random forest based on the percentage increases in mean square error (%IncMSE) and node purity (%IncNodePurity). Ethanol concentration was a more important variable with a higher %IncMSE and %IncNodePurity than contact time for the inactivation of enveloped and non-enveloped viruses with the available organic matter. Because specific guidelines for virus inactivation by ethanol are lacking, data analysis using machine learning is essential to gain insight from certain datasets. We provide new knowledge for determining guideline values related to the selection of ethanol concentration and contact time that effectively inactivate viruses.