Predictive Toxicology of Chemicals: Experiences and Impact of AI Tools
Papers from the AAAI Spring Symposium
Giuseppina C. Gini and Alan R. Katritzky, Cochairs
AI and related techniques play a major role in toxicity prediction. The goal of computational toxicity prediction is to describe the relationship between chemical properties, on the one hand, and biological and toxicological processes, on the other. This symposium will highlight the potential of different AI approaches, either individually and combined, for computational toxicity prediction.
Success in this research depends on the contribution of experts from different areas, and we invite participation from researchers in all related fields. We welcome AI researchers who have applied learning techniques to domains outside toxicity prediction and are in search of new areas. Some of the questions to be addressed in the symposium follow.
How do we represent chemical information? Several methods have been proposed. Are they equivalent and useful for toxicity prediction? How do we evaluate them? Are results from different experiments reproducible? A tutorial on this topic will be hold. A series of possibilities will be illustrated in different presentations.
How can machine learning techniques be used? AI tools have yet to be fully evaluated in this domain. Which techniques are better for toxicity prediction, especially given our changing understanding of toxicology? Are hybrid approaches better? A tutorial on the state-of-the-art of computer-based methods will be held. ILP, argumentation, ANN, Bayesian methods, fuzzy logic, mathematical and statistical methods, such as discriminant analysis, lattice theory and hybrid systems will be presented and discussed in many communications.
Are current experimental data sets sufficient for AI techniques? Do they have sufficient accuracy? How do we take advantage of existing data sets? Can we use techniques from data mining and reasoning under uncertainty?
Experiences from international projects will be presented. Needs and future trends will be discussed, giving voice to end-users of these software systems to achieve an integration of research and real-world applications.