VALIDATION OF PREDICTION OF NOAEL AND LOAEL USING FIVE DIFFERENT MODELS IN VEGA SOFTWARE

• Main Author: Nurrahmah, Aulia
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/82625
• Institution: Institut Teknologi Bandung

During the drug development stage, in vivo preclinical trials results in No Observed Adverse Effect Level (NOAEL) and Lowest Observed Adverse Effect Level (LOAEL). NOAEL and LOAEL data obtained can be used to determine the initial dose, namely the dose given in the first clinical trials in humans. However, in vivo testing requires large costs and considerable time. Therefore, sustainable efforts have emerged to predict toxicity in silico. A total of 3 QSAR models for NOAEL prediction and 2 LOAEL prediction models are available for free in the VEGA software. Humans are exposed to chemicals in food and the environment, and one of the goals of chemical risk assessment is to establish safe levels to protect public health from chronic exposure. Therefore, in this research, data was segmented into groups of cosmetics, food additives, medicines, household supplies, pesticides and dyes. This research aims to validate the NOAEL and LOAEL models using several statistical parameters such as correlation coefficient (R2), concordance correlation coefficient (CCC), index of ideality of correlation (IIC), mean absolute error (MAE), and root mean square error (RMSE) while determining which group of compounds giving the best statistical parameters so that they can be predicted using any available models. The food additives segment results in the best statistical parameters meaning it can be predicted using the models available in VEGA. Suggestions for further model development are to create models per segmentation so that they can be more reliable for specific purposes.