Previsão da solubilidade de fármacos orgânicos em água, baseado na relação quantitativa entre estrutura molecular e propriedades in silico

• Ano de defesa: 2010
• Autor(a) principal: Peixoto, Edilson Pinheiro
• Orientador(a): Camargo, Ademir João
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Estadual de Goiás
• Programa de Pós-Graduação: Programa de Pós-Graduação Stricto sensu em Ciências Moleculares
• Departamento: UEG ::Coordenação de Mestrado Ciências Moleculares
• País: Brasil
• Palavras-chave em Português: Solubilidade; Previsão solubilidade em água; QSPR; Propriedades moleculares in silico
• Palavras-chave em Inglês: Solubility; Water solubility prediction; QSPR; Molecular proprierties in silico
• Área do conhecimento CNPq: CIENCIAS EXATAS E DA TERRA::QUIMICA
• Link de acesso: http://www.bdtd.ueg.br/handle/tede/503

The solubility of a substance, with potential to become a comercial drug, in water is of extreme importance to the development of the drug itself. In this study the in silico topological, geometric and electronic proprieties of 322 molecules, structurally different from each other, were calculated using their 3D spatial structure. Among the proprieties calculated we considered: molecule area, volume, ionization potential, “HOMO”, “LUMO”, dipolar moment and polarizability (alfa), using a semi-empiric model (RM1), and also the number of hydrogen bond donators and acceptors, obtained from the literature. The data was divided into 2 groups, one with 242 molecules, for the development of a predictive model for molecular solubility, and another, with 80 molecules for the model validation. The model data was tested for correlation with the logarithm of the solubility, and a multi-linear regression model was developed, so to achieve an equation capable of predicting the solubility with only in silico proprieties. This equation resulted in the parameters (r)=0,716 and (R2 )=51,3%, with S=1,19463, referring to logS, relatively larger than those found in literature. The calculated logS for the 80 validation molecules showed the predictive potential of the model, resulting in a correlation of (r)=0,751, and relative mean error of 4,5%, when compared to the experimental logS.