Identification of the descriptor pharmacophores using variable selection QSAR: applications to database mining.

Area of Science:

• Computational Chemistry
• Medicinal Chemistry
• Drug Discovery

Background:

• The pharmacophore concept is crucial for rational drug design, traditionally defined by 3D arrangements of functional groups in active molecules.
• Existing methods often rely on complex 3D structural information for compound activity prediction.

Purpose of the Study:

• To introduce and validate a generalized concept of 'descriptor pharmacophore' for drug discovery.
• To enhance the efficiency of chemical database mining for identifying biologically active compounds.

Main Methods:

• Development and application of two variable selection quantitative structure-activity relationship (QSAR) methods: Genetic Algorithms-Partial Least Squares (GA-PLS) and K-Nearest Neighbors (KNN).
• Utilizing multiple topological molecular descriptors, including molecular connectivity indices and atom pairs (AP).
• Employing stochastic optimization algorithms to build robust QSAR models with high cross-validated R2 (q2) values.

Main Results:

• Descriptor pharmacophores, defined by QSAR-selected subsets of molecular descriptors, establish statistically significant structure-activity correlations.
• These descriptor pharmacophores represent invariant selections of descriptor types, facilitating efficient similarity searches.
• Chemical similarity searches using descriptor pharmacophores are more efficient than using all descriptors for mining databases and virtual libraries.

Conclusions:

• The descriptor pharmacophore concept offers a powerful alternative to traditional pharmacophores in rational drug design.
• This approach significantly improves the efficiency of discovering compounds with desired biological activities through enhanced database mining.