BeeToxAI: Um Portal Online Baseado em Inteligência Artificial para Predizer a Toxicidade Aguda de Produtos Químicos em Abelhas Melíferas

Favorite this paper
How to cite this paper?
Details
  • Presentation type: Exposição de Pôster
  • Track: Química Ambiental - AMB
  • Keywords: Cheminformatics; ecotoxicology; Artificial Intelligence; Predictive modeling; Pesticides;
  • 1 LabMol – Laboratory for Molecular Modeling and Drug Design, Faculty of Pharmacy, Universidade Federal de Goiás, Goiás, 74605-170, Brazil
  • 2 InsilicAll Inc., São Paulo, 04363-090, Brazil
  • 3 Laboratory for Molecular Modeling, UNC Eshelman School of Pharmacy, University of North Carolina at Chapel Hill, North Carolina, 58059-900, USA
  • 4 Universidade Federal de Goiás

BeeToxAI: Um Portal Online Baseado em Inteligência Artificial para Predizer a Toxicidade Aguda de Produtos Químicos em Abelhas Melíferas

Bruno Neves

UNIVERSIDADE FEDERAL DE GOIÁS

Abstract

The use of pesticides has been associated with the drastic decline of bee populations in several countries. Honey bee species such as Apis mellifera are of utmost importance because of their vital ecological duty as pollinators of wild plants and crops. Consequently, scientific advisory bodies have standardized in vivo protocols to access acute toxicity of chemicals in honey bees. In view of the costly and time-consuming characteristics of experimental assays, we developed an innovative artificial intelligence-based web portal to access acute toxicity of chemicals in honey bees. Initially, we have collected and curated the largest publicly available dataset of compounds with acute contact and oral toxicity for A. mellifera. Then, 40 machine learning (ML) models were built using contact and oral datasets to distinguish toxic (LD50 ≤11 µg/bee) vs. non-toxic (LD50 >11 µg/bee) compounds. The model built using FeatMorgan_2 fingerprint + Random Forest demonstrated the best external predictivity among all other models developed for acute contact toxicity (ACC = 92%; SE = 83%; SP = 96%; and MCC = 81%). The model built using MACCS fingerprints + Support Vector Machine shown the best external predictivity among all other models developed for acute oral toxicity (ACC = 91%; SE = 93%; SP = 93%; and MCC = 82%). These models are freely available through the BeeToxAI web portal at http://beetoxai.labmol.com.br/. The BeeToxAI (Figure 1) allows users for fast screening of compounds. After drawing or importing a molecule and submitting it to a prediction, the user will receive in few seconds the predicted outcomes for acute contact and oral toxicities. These predictions are followed by probability estimates, which are useful for estimating the confidence of predicted outcomes, the applicability domain estimates, and mechanistic interpretation of prediction using contribution maps (atoms with positive and negative contributions for toxicity). In summary, BeeToxAI can be used as a reliable alternative to animal tests for predicting acute toxicity of chemicals in honey bees.

Questions (1 topic)

Share your ideas or questions with the authors!

Did you know that the greatest stimulus in scientific and cultural development is curiosity? Leave your questions or suggestions to the author!

Sign in to interact

Have a question or suggestion? Share your feedback with the authors!

Author

Bruno Neves

Olá Luis Fernando

Fico feliz que tenha gostado do trabalho. Nosso grupo de pesquisa aplica extensivamente AI (machine learning, deep learning) para projetos de descoberta de fármacos e predição de propriedades toxicológicas de ingredientes de cosméticos, agrotóxicos, protótipos de fármacos e produtos de degradação de medicamentos. Todas as ferramentas são desenvolvidas para atender não somente projetos de inovação, mas também aspectos regulatórios (registro e reavaliação).

O BeeToxAI foi o primeiro aplicativo voltado para avaliação preditiva de toxicidade de agrotóxicos. Atualmente, estamos desenvolvendo aplicativos semelhantes para minhocas, aves, peixes, microcrustáceos e algas, visando atender todos os níveis tróficos de ecosistemas.