Importance of developing new descriptors in QSAR/QSPR (2pag)

DAMIAN PALOMBA; GUSTAVO E. VAZQUEZ; MONICA F. DIAZ

Rosario de Santa Fe

Congreso; 4to. Congreso Argentino de Bioinformática y Biología Computacional (4CAB2C); 2013

CIFASIS

Many researchers have adopted computational approaches to predict different behaviors, from biological activity to physicochemical properties in drugs, materials, etc. In particular, the QSAR/QSPR (Quantitative Structure-Activity/Property Relationship) technique relates specific parameters of molecule structure (descriptors) to the studied activity/property by using a dataset of molecular entities and experimental target values. One of the most important aspects to take into account when developing QSAR/QSPR models is to employ descriptors that truly describe the relevant aspects of the molecule. Traditionally, prediction models are built by using well-known parameters, such as 0-, 1- 2- 3-D descriptors. However in some applications these molecular parameters cannot grasp characteristics related to the synthesis, processing, or sample preparation. One of these situations is the modeling of mechanical properties of polymers in bio and pharma technologies (biopolymers, polymeric capsules in drug delivery, etc). Furthermore, by addressing the problematic of the synthetic polymers, the difficulty of the molecular design occurs since the depiction of polymeric structures cannot be clearly defined in contrast to the small molecules. Among other factors to consider are: the structural (e.g., chain length, tacticity, and monomer segments) and the composition ones (monomer content, blends, and additives). In this work we present a comparison between different prediction models for glass-transition temperature (Tg) and elongation at break by using general purpose descriptors and novel ad-hoc developed descriptors.