Quantitative Structure−Activity Relationship of Flavonoid p56^lck Protein Tyrosine Kinase Inhibitors. A Neural Network Approach

Abstract

Specific inhibitors of protein tyrosine kinase as antiproliferative agents are instrumental in several aspects of neoplastic disease and have found wide interest as potential pharmacological agents. We have applied an artificial neural network based on a counterpropagation algorithm to develop quantitative structureactivity relationships in a large dataset of 105 flavonoid derivatives that inhibit the enzyme p56 lck protein tyrosine kinase. The results of such approach were compared with the linear multiregression analysis with regard to the ability to fit biological activity surfaces, predict activity, and explore the nonlinear aspects of the dependence of activity on properties. Excellent correlation was obtained for both classical and quantum chemical descriptors, and relevance of the descriptors to binding properties of the enzyme receptor active site is hypothesized.

The Effect of Variable Selection on the Non‐linear Modelling of Oestrogen Receptor Binding

“…It should be noted however, that reducing the number of epochs does not improve the fit for the test set, although it reduces the fit for the training set. The number of epochs examined in this study is within the range used in a number of other previous studies [35,36,46,47]. An analysis of the outliers showed that some of the compounds are mis-predicted in most of the models.…”

“…The numbers of epochs were between 80 and 220 with steps of 10, followed by 250 and 300 epochs. Other computational parameters of the networks were the maximal and minimal learning rates which were set at 0.5 and 0.01, non-thoroidal boundary conditions, triangular correction function of the neighbourhood (as recommended in references [35,36]). …”

Application of a Pruning Algorithm To Optimize Artificial Neural Networks for Pharmaceutical Fingerprinting

“…The current approach can be extended for detection of localized fingerprinting regions and interpretation of results is neural network analysis of mass spectra 19 and infrared spectra 20,21 or for QSAR (quantitative structure-activity relationship) studies using, for example, "spectrumlike" representations of chemical structures. 22 The general idea of pruning can be also used for interpretation of more complex ANNs, such as the neural device proposed by Bashkin et al 23 There are certain advantages gained by decreasing the width of the fingerprint region. For example, the familiar drift of HPLC signals is known to increase with retention time due to the physical conditions of the chromatographic experiment (temperature, pressure, etc.)…”

Toward an Optimal Approach for Variable Selection in Counter‐Propagation Neural Networks: Modeling Protein‐Tyrosine Kinase Inhibitory of Flavanoids Using Substituent Electronic Descriptors

“…This data set was already used in some QSAR studies. [30] The measured IC 50 (mM) is defined as the molar concentration of the flavonoids necessary to give half-maximum inhibition. As it is shown in Table 1, 55 compounds (52.4 % of the whole data) represented log(1/IC 50 ) > 2.7 and the rest (47.6 %) possessed activity values equal or lower than 2.7.…”