N-tuple topological/geometric cutoffs for 3D N-linear algebraic molecular codifications: variability, linear...

Area of Science:

• Computational chemistry
• Cheminformatics
• Quantitative Structure-Activity Relationship (QSAR) studies

Background:

• The QuBiLS-MIDAS framework utilizes molecular descriptors (MDs) to represent molecular structures.
• Existing methods (QuBiLS-MIDAS KA-MDs) consider all inter-atomic relations, potentially including noise.
• There is a need for more selective methods to capture relevant atomic interactions.

Purpose of the Study:

• To introduce novel N-tuple topological/geometric cutoffs for the QuBiLS-MIDAS framework.
• To develop new molecular descriptors (QuBiLS-MIDAS NQ-MDs) based on these cutoffs.
• To evaluate the performance of the new descriptors in QSAR studies.

Main Methods:

• Definition of kth two-, three-, and four-tuple topological and geometric neighborhood quotient (NQ) matrices.
• Analysis of chemical space to determine optimal cutoff intervals for distances, angles, and volumes.
• Application of Shannon's entropy for variability analysis and Principal Component Analysis (PCA) for information codification.
• Development of QSAR models using multiple linear regression on benchmark datasets (steroids, ACE, thermolysin, thrombin inhibitors).

Main Results:

• The proposed N-tuple cutoffs (QuBiLS-MIDAS NQ-MDs) show better distribution patterns compared to 'Keep All' (KA-MDs).
• PCA indicates that NQ-MDs capture both existing and novel chemical information.
• QSAR models based on NQ-MDs demonstrate superior predictive performance over KA-MDs across multiple datasets.
• The novel cutoffs effectively codify specific inter-atomic relations.

Conclusions:

• The N-tuple topological/geometric cutoffs are a valuable criterion for generating molecular descriptors.
• These novel descriptors enhance the modeling capacity of the QuBiLS-MIDAS 3D-MDs.
• The proposed methods offer improved accuracy in predicting biological activities through selective consideration of atomic relations.