Developing ontologies for improving and interpreting regression models applied to Quantitative Structure-Activity Relationship studies

• Grant number: 18/12374-6
• Support Opportunities: Scholarships abroad - Research
• Effective date (Start): March 04, 2019
• Effective date (End): March 03, 2020
• Field of knowledge: Physical Sciences and Mathematics - Computer Science - Computing Methodologies and Techniques
• Principal Investigator: Patrícia Rufino Oliveira
• Grantee: Patrícia Rufino Oliveira
• Host Investigator: James Geller
• Host Institution: Escola de Artes, Ciências e Humanidades (EACH). Universidade de São Paulo (USP). São Paulo , SP, Brazil
• Research place: New Jersey Institute of Technology (NJIT), United States

Abstract

Quantitative Structure-Activity Relationship (QSAR) studies are crucial to the drug discovery process as these analyses are mainly used for estimating values that indicate the biological activity of new chemical compounds. However, these studies are very costly, since the transformation of chemical descriptors into numerical data sets is not a trivial task. In such situations, it would be desirable to use information from other sets of existing chemical data for improving the performance of a related QSAR model built from a very small training data set. In order to address this issue, transfer learning techniques could be applied, reducing the efforts of researchers and the cost of the process to generate new sets of chemical descriptors. The development of such approaches was primarily motivated by the fact that people can apply the knowledge acquired previously to solve new problems more quickly and with better solutions.A number of machine learning methods have been successfully applied for constructing QSAR models, although the use of such approaches is critically dependent on the availability of large enough data sets to be used in the training phase and methods that enable interrogation and interpretation. Consequently, one also believes that by combining the generalized predictive power of machine learning methods and knowledge extracted from biological and chemical ontologies can lead to important contributions to researches in artificial intelligence and medicinal chemistry areas. This research project aims at developing and applying ontological data to guide the selection of suitable (related) data sets for transfer knowledge from one QSAR model to another. Furthermore, an additional objective refers to use the ontological data to explore the drug-target relations estimated by the constructed QSAR models in order to understand how the modifications on structural characteristics of compounds induce their biological activity. By combining chemical and biological components and their interrelationships in an integrated ontology for QSAR studies, the project will provide a more holistic and standardized knowledge formalization of the drug discovery process. One also expects that such resulting ontological data can be used to guide the selection of source data sets in QSAR transfer learning applications and to interpret predictive results obtained by machine learning methods in medicinal chemistry area.