Abstract
The prediction of physicochemical and biochemical endpoints related to food, cosmetics, drug discovery, and medicine using computational approaches is relevant to the need of the modern society. The experimental determination of the endpoints is impossible for all necessary substances.
The Index of Ideality of Correlation (IIC) is a criterion of predictive potential of a model. The index has been implemented in the CORAL software, which is a tool to build up quantitative structureactivity relationships.
In the present work, two models for sweetness potential are built using this software. The first model is built without taking into account the IIC.
The second model is built with taking into account the IIC. The results show that the IIC improves the predictive potential.
This was confirmed in three computational experiments with random splits of the data related to sweetness into training and validation sets. [GRAPHICS] .