Naslov
The extra-trees regressor model in predicting autonomic neurotoxicity

Autori
Varnai, Veda Marija ; Kovačić, Jelena ; Babić, Željka

Vrsta, podvrsta i kategorija rada
Sažeci sa skupova, sažetak, znanstveni

Skup
6th Croatian congress of toxicology with international participation (CROTOX 2021)

Mjesto i datum
Rabac, Hrvatska, 03.10.2021. - 06.10.2021

Vrsta sudjelovanja
Poster

Vrsta recenzije
Recenziran

Ključne riječi
autonomic nervous system ; in silico ; intraclass correlation coefficient ; LD50 ; mice

Sažetak
Existing in silico models for predicting neurotoxicity of chemicals still have numerous limitations. Among the recently published models, the extra-trees regressor (ETR) model was recommended by Jiang et al. (2020) for quantitative prediction of autonomic nervous system toxicity. Our study aimed to assess the applicability of this model, developed on mouse data after intraperitoneal application of various chemicals, on the same species exposed via intravenous (i.v.), subcutaneous (s.c.) and peroral route (p.o.), for the same neurotoxic endpoint. Mouse LD50 data related to autonomic neurotoxicity were collected from the ChemIDplus database by webchem R package. Molecular descriptors were calculated by PyBioMed package for 86 organic chemicals for i.v., 65 for s.c. and 100 for p.o. routes. A separate ETR model for each route was trained to predict pLD50 [- log10(LD50 mol/kg)] related to autonomic neurotoxicity, using 19 molecular descriptors selected as most suitable by Jiang et al., and evaluated by leave-one-out cross-validation. Good agreement between the predicted pLD50 values and those stated in the ChemIDplus database was found for the s.c. route (intraclass correlation 0.82, 95 % confidence interval 0.72- 0.89, P<0.001), although root mean square error was three times higher than in the model by Jiang et al.. Moderate agreement was observed for the i.v. route (ICC 0.64, 95 % CI 0.50-0.75, P<0.001), and poor for the p.o. route (ICC 0.41, 95 % CI 0.23-0.56, P<0.001). The results indicate that the model developed by Jiang et al. has limited applicability for other exposure routes, especially non-parenteral, even when model parameters are learned specifically for a respective exposure route in the same species.

Izvorni jezik
Engleski

Znanstvena područja
Javno zdravstvo i zdravstvena zaštita

POVEZANOST RADA