ABSTRACT
Families of molecular descriptors (FMD) were designed and used to relate the structural information with measured properties and activities for different series of chemical compounds. The strategy to materialize a family of molecular descriptors is to use a genetic code embedding in it a series of different operators applying on the structure of the chemical compound to produce a parameterized numerical outcome. Modeling the molecular structure is the way of understanding of the microscopic level and its expression at the macroscopic one level. Modeling the molecular structure is a prerequisite for structure-activity inference analysis. All implementations were parallelized for exploiting the multi-core feature of the modern computers. Parallelization is portable when is implemented to pass to different tasks different molecules from a set. The using of FMDs to estimate and predict biological activities proved constantly better results when was compared with other alternatives.
