AN INTEGRAL FRAMEWORK FOR QSAR MODELLING USING COMPUTATIONAL INTELLIGENCE AND VISUAL ANALYTICS

FIORELLA CRAVERO; MARÍA JIMENA MARTÍNEZ; GUSTAVO E. VAZQUEZ; MÓNICA. F. DIAZ; IGNACIO PONZONI

Bahía Blanca

Congreso; VI Argentinian Conference on Bioinformatics and Computational Biology; 2015

AB2C2 Y (ICIC) Instituto de Ciencias e Ingeniería de la Computación - UNS- CONICET

Historically, the development of new polymeric materials consisted of a trial and error process, which involved the synthesis of new products in order to characterize experimentally and only then qualify it according to the desired application. Currently, the demand of the market for pre-specified properties leads to polymers industry to seek lower cost of synthesis using computational methods of prediction of properties, i.e. forecast in silico the estimated behaviour of a material of design, prior to its synthesis. This task is not simple; in addition to having a high molecular weight and the polydisperse structure, molecular representations are very synthetic and they deviate from reality [1]. In addition the emerging of modern technologies allowed the rapid chemical synthesis and testing at high speed of large collections of compounds giving a significant growth of databases with heterogeneous data of different experiments. This new available data led to the use of machine learning techniques, data mining and statistical tools to discover new patterns and structures that serve to infer knowledge on the relationship between the chemical structure of a compound and its physicochemical properties. In this way, through the development and use of these computational methods, the aim is to identify and prioritize new candidate materials with special characteristics prior to its synthesis, thus saving the high costs associated with the design process [2,3].