Influence of Descriptor Implementation on Compound Ranking Based on Multiparameter Assessment

Area of Science:

• Medicinal Chemistry
• Computational Chemistry
• Drug Discovery

Background:

• Multiple implementations exist for common molecular descriptors, potentially affecting compound prioritization.
• Multiparameter assessment (MPA) is crucial for balancing diverse molecular and pharmacological properties in drug design.

Purpose of the Study:

• To evaluate the interchangeable use of different molecular descriptor implementations in MPA.
• To assess the impact of calculation methods, data distribution, applicability domains, and desirability functions on MPA compound ranking stability.

Main Methods:

• Analysis of common descriptors (logP, logS, TPSA, logBB, hERG, nHBA) across various implementations.
• Examination of descriptor calculation methods, continuity/discreteness, applicability domains, and desirability functions.
• Stability assessment of MPA compound rankings based on descriptor implementation variations.

Main Results:

• Interchangeable use of descriptor implementations is acceptable only for continuous parameters with smooth desirability functions.
• Discrete parameters require using the exact descriptor implementation used for the MPA scoring profile.
• Inconsistent compound assessments arise from differing descriptor applicability domains.

Conclusions:

• Careful selection of molecular descriptor implementations is vital for reliable MPA in drug optimization.
• The nature of descriptor data (continuous vs. discrete) dictates the feasibility of using different implementations.
• Understanding descriptor applicability domains is essential to avoid assessment inconsistencies in MPA.