| Summary: | 碩士 === 國立東華大學 === 化學系 === 97 === Aromatic amines often found in various industrials and environments are of great risk to humans due the chronic exposure to these toxic substances, which may cause mutagenicity or carcinogenicity in the worst case. It is of critical importance to develop an in silico model to fast predict the mutagenicity of aromatic amines found in environmental pollutants and industrial products. This study used hierarchical support vector regression, support vector machine and genetic function approximation machine learning techniques based on the mutagenicity data (Ames test TA100) for various aromatic amines comprehensively complied from the literature. The predictions by these four machine learning models are in agreement with the experimental observations for those molecules in the training set, the test and the outlier set. Various statistical analyses and validation criteria asserted their accuracy and predictivity. They also performed better than their traditional linear QSAR counterparts by different cross-comparisons. Of these four machine learning models, the HSVR outperformed the other three models in almost every aspect. Thus, this accurate, predictive and robust HSVR model can be employed as a tool to predict mutagenicity of new aromatic amines.
|
|---|
