(56a) Computational Approach of Thermodynamic Fragmentation Applied to Formulating Badly Soluble Actives

We address one of the major computational problems in the pharmaceutical and agrochemical business: the rational design of delivery vehicles for badly soluble actives. For example, in a typical scenario in a discovery/development cycle, one has only a few months, or even weeks, to find a proper formulation, driven by very costly time expenditure way down the pipeline. The chemical space for formulations is enormous, and even if one considers modern high-throughput experimentation, chances are the optimal formulation cannot be found by experimentation alone. For chemical informatics, such design also poses quite some challenges, since classical descriptor technology is almost always unitary (component specific), and therefore not suited for the non-linear binary interactions between active and delivery matrix. The more so, since in many cases one seeks or tests delivery systems that are based on some self-assembly structure, such as an emulsion, micelle or liposome. In contrast, in existing chemical informatics methods for the much simpler logP predictions one can rely on such unitary descriptors quite well, since the (iso-octanol/water) matrix in this case is known and constant over a known training set. Our strategy is completely different. We develop computational screening technologies that do take into account the structural resolution of both active and matrix, by employing a novel concept of thermodynamic fragmentation. Thereby the interactions of drug-matrix is described as a interaction between fragments, with the advantage that the fragmentation can be calibrated by (non-pharma) engineering databases. The presentation discusses in depth the motivation, the algorithm and some highlights from recent applications.