Odor threshold prediction by means of the Monte Carlo method

• A predictive model for odor threshold is developed.
• The statistical quality of the model is quite good.
• Simplified molecular input-line entry system is used to represent molecular structure.
• The CORAL software is utilized to build up the model.
• The model is built up in accordance with OECD principles.

A large set of organic compounds (n=906) has been used as a basis to build up a model for the odor threshold (mg/m3). The statistical characteristics of the best model are the following: n=523, r2=0.647, RMSE=1.18 (training set); n=191, r2=0.610, RMSE=1.03, (calibration set); and n=192, r2=0.686, RMSE=1.06 (validation set). A mechanistic interpretation of the model is presented as the lists of statistical promoters of the increase and decrease in the odor threshold.

Figure optionsDownload as PowerPoint slide