Prediction of the GC-MS Retention Indices for a Diverse Set of Terpenes as Constituent Components of Camu-camu (Myrciaria dubia (HBK) Mc Vaugh) Volatile Oil, Using Particle Swarm Optimization-Multiple Linear Regression (PSO-MLR)

Abstract

A reliable quantitative structure retention relationship (QSRR) study has been evaluated to predict the retention indices (RIs) of a broad spectrum of compounds, namely 118 non-linear, cyclic and heterocyclic terpenoids (both saturated and unsaturated), on an HP-5MS fused silica column. A principal component analysis showed that seven compounds lay outside of the main cluster. After elimination of the outliers, the data set was divided into training and test sets involving 80 and 28 compounds. The method was tested by application of the particle swarm optimization (PSO) method to find the most effective molecular descriptors, followed by multiple linear regressions (MLR). The PSO-MLR model was further confirmed through “leave one out cross validation†(LOO-CV) and “leave group out cross validation†(LGO-CV), as well as external validations. The promising statistical figures of merit associated with the proposed model (R2train=0.936, Q2LOO=0.928, Q2LGO=0.921, F=376.4) confirm its high ability to predict RIs with negligible relative errors of predictions (REP train=4.8%, REP test=6.0%).