QSAR as a random event: a case of NOAEL
2015
Toropova, Alla P. | Toropov, Andrey A. | Veselinović, Jovana B. | Veselinović, Aleksandar M.
Quantitative structure–activity relationships (QSAR) for no observed adverse effect levels (NOAEL, mmol/kg/day, in logarithmic units) are suggested. Simplified molecular input line entry systems (SMILES) were used for molecular structure representation. Monte Carlo method was used for one-variable models building up for three different splits into the “visible” training set and “invisible” validation. The statistical quality of the models for three random splits are the following: split 1 n = 180, r ² = 0.718, q ² = 0.712, s = 0.403, F = 454 (training set); n = 17, r ² = 0.544, s = 0.367 (calibration set); n = 21, r ² = 0.61, s = 0.44, r ₘ ² = 0.61 (validation set); split 2 n = 169, r ² = 0.711, q ² = 0.705, s = 0.409, F = 411 (training set); n = 27, r ² = 0.512, s = 0.461 (calibration set); n = 22, r ² = 0.669, s = 0.360, r ₘ ² = 0.63 (validation set); split 3 n = 172, r ² = 0.679, q ² = 0.672, s = 0.420, F = 360 (training set); n = 19, r ² = 0.617, s = 0.582 (calibration set); n = 21, r ² = 0.627, s = 0.367, r ₘ ² = 0.54 (validation set). All models are built according to OCED principles.
显示更多 [+] 显示较少 [-]