New QSAR combined strategy for the design of A1 adenosine receptor agonists.

Combined discriminant and regression analysis was carried out on a series of 167 A1 adenosine receptor agonists to identify the best linear and nonlinear models for the design of new compounds with a better biological profile. On the basis of the best linear discriminant analysis and both linear and nonlinear Multi Layer Perceptron neural networks regression, we have designed and synthesized 14 carbonucleoside analogues of adenosine. Their biological activities were predicted and experimentally measured to demonstrate the capability of our model to avoid the prediction of false positives. A good agreement was found between the calculated and observed biological activity.

BCUT descriptors to predicting affinity toward A3 adenosine receptors.

The BCUT descriptors have been applied to the study of the A(3) adenosine receptor agonist effect of 32 adenosine analogues. A model, able to describe more than 80% of the variance in the experimental activity was developed with the use of the above-mentioned approach. Four different approaches (topological, Galvez topological charges indexes, Randic molecular profiles, and geometrical descriptors) failed to give satisfactory models for this property with the same number of variables in the equation. Although statistically significant models were derived containing descriptors other than BCUT, the best fitted model was still found with these descriptors.

Geometry, topology, and atom-weights assembly descriptors to predicting A1 adenosine receptors agonists.

The GEometry, Topology, and Atom-Weights AssemblY (GETAWAY) approach has been applied to the study of the A1 adenosine receptors agonist effect of 32 adenosine analogues: N6-arylcarbamoyl, 2-arylalkynyl-N6-arylcarbamoyl, and N6-carboxamido derivatives. A model, able to describe more than 77% of the variance in the experimental activity, was developed with the use of the above mentioned approach. Five different approaches (Topological, Galvez Topological Charges indexes, Randic Molecular Profiles, Geometrical, and WHIM descriptors) failed to give satisfactory models (R2=0.70) for this property with the same number of variables in the equation. Although statistically significant models were derived containing descriptors other than GETAWAY, the best fitted out model was still found with these descriptors.

A radial distribution function approach to predict A(2B) agonist effect of adenosine analogues.

The radial distribution function (RDF) approach has been applied to the study of the A(2B) agonist effect of a set of 89 adenosine analogues reported with this activity. A model able to describe more than 70% of the variance in the experimental activity was developed with the use of the mentioned approach. In contrast, none of the eleven different approaches including the use of Constitutional, Topological, Molecular walk count, BCUT, Galvez topological charge indices, 2D autocorrelations, Randic molecular profiles, Geometrical, 3D Morse, WHIM and GETAWAY descriptors was able to explain more than 47% of the variance in the mentioned property with the same number of descriptors.