Self-organizing maps and VolSurf approach to predict aldose reductase inhibition by flavonoid compounds

ABSTRACT

Aldose Reductase (AR) is the polyol pathway key enzyme which converts glucose to sorbitol. High glucose availability in insulin resistant tissues in diabetes leads into an accumulation of sorbitol, which has been associated with typical chronic complications of this disease, such as neuropathy, nephropathy and retinopathy. In this study, 71 flavonoids AR inhibitors were subjected to two methods of SAR to verify crucial substituents. The first method used the PCA (Principal Component Analysis) to elucidate physical and chemical characteristics in the molecules that would be essential for the activity, employing VolSurf descriptors. The rate obtained explained 53 percent of the system total variance and revealed that a hydrophobic-hydrophilic balance in the molecules is required, since very polar or nonpolar substituents decrease the activity. Artificial Neural Networks (ANNs) was also employed to determine key substituents by evaluating substitution patterns, using NMR data. This study had a high success rate (85 percent accuracy in the training set and 88 percent accuracy in the test set) and showed polihydroxilations are essential for high activity and methoxylations and glicosilations primarily at positions C7, C3' and C4' decrease the activity.