2D Quantitative structure-property relationship study of mycotoxins by multiple linear regression and support vector...

Area of Science:

• Analytical Chemistry
• Computational Chemistry
• Toxicology

Background:

• Mycotoxins pose significant risks to food safety and human health.
• Accurate prediction of mycotoxin retention times is crucial for chromatographic analysis.
• Quantitative Structure-Property Relationship (QSPR) studies offer a computational approach to predict chemical properties.

Purpose of the Study:

• To develop and validate QSPR models for predicting retention time (t(R)) of 67 mycotoxins.
• To identify significant molecular descriptors influencing mycotoxin retention in liquid chromatography-UV-mass spectrometry.
• To compare the performance of Support Vector Machines (SVMs) and Multiple Linear Regression (MLR) for QSPR modeling.

Main Methods:

• Calculation of molecular descriptors from optimized 3D structures of 67 mycotoxins.
• Application of variable selection techniques including missing value, zero, and multicollinearity tests, and genetic algorithms.
• Development of QSPR models using SVMs and MLR.
• Statistical validation including correlation coefficient (r^2), predictability (q^2), and applicability domain (AD) assessment using William's plot.

Main Results:

• Both SVM and MLR models demonstrated high accuracy in predicting mycotoxin retention times.
• SVM models achieved r^2 = 0.931 and q^2 = 0.932, while MLR models achieved r^2 = 0.923 and q^2 = 0.915.
• The models were robust and their applicability domains were well-defined, indicating reliable predictions.

Conclusions:

• The developed QSPR models effectively predict mycotoxin retention times using molecular descriptors.
• SVMs and MLR are suitable computational techniques for QSPR studies in mycotoxin analysis.
• These models can serve as valuable tools for chromatographic method development and mycotoxin identification.