Bench to bedside: The ambitious goal of transducing medicinal chemistry from the lab to the clinic

This paper deals with a critical examination on the possibility of quantitatively predicting the in vivo activity of new chemical entities (NCEs) by making use of in silico and in vitro data including three-dimensional structure of drug-target complex, thermodynamic and crowding parameters, ADME (absorption, distribution, metabolism, excretion) properties, and off-target (toxic) interactions. This formidable challenge is still a dream, given the presently occurring exceedingly high (>95%) attrition rates of NCEs. As a solution we envisage exploiting advanced AI (artificial intelligence) algorithms. In fact, very recent AI implemented programs proved remarkably effective and accurate in predicting the 3D architecture of (any) protein, starting from the amino-acid sequence only. The same accuracy could not be obtained using classical conformational studies. Apart from these breakthrough results, AI algorithms could be profitably used to extract valuable information from the huge amount of data so far accumulated from previous studies. In case of positive results, the drug discovery procedure would be sensibly accelerated, and the relative costs remarkably reduced.