| Abstract
| - We have performed virtual screening using some very simple features, by employing the number of atomsper element as molecular descriptors but without regard to any structural information whatsoever. Surprisingly,these atom counts are able to outperform virtual-affinity-based fingerprints and Unity fingerprints in someactivity classes. Although molecular weight and other biases were known in target-based virtual screeningsettings (docking), we report the effect of using very simple descriptors for ligand-based virtual screening,by using clearly defined biological targets and employing a large data set (>100 000 compounds) containingmultiple (11) activity classes. Structure-unaware atom count vectors as descriptors in combination with theEuclidean distance measure are able to achieve “enrichment factors” over random selection of around 4(depending on the particular class of active compounds), putting the enrichment factors reported for moresophisticated virtual screening methods in a different light. They are also able to retrieve active compoundswith novel scaffolds instead of merely the expected structural analogues. The added value of many currentlyused virtual screening methods (calculated as enrichment factors) drops down to a factor of between 1 and2, instead of often reported double-digit figures. The observed effect is much less profound for simpledescriptors such as molecular weight and is only present in cases of atypical (larger) ligands. The currentstate of virtual screening is not as sophisticated as might be expected, which is due to descriptors still notbeing able to capture structural properties relevant to binding. This fact can partly be explained by highlynonlinear structure−activity relationships, which represent a severe limitation of the “similar propertyprinciple” in the context of bioactivity.
|
