QuBiLs-MAS

QuBiLs-MAS computes topological molecular descriptors from 0D to 2.5D using graph-theoretic electronic-density and edge-adjacency matrices together with bilinear, quadratic, and linear algebraic forms to characterize atom- and bond-based molecular features.

Key Features:

• Algebraic Forms and Matrix Transformations: Calculates molecular descriptors using bilinear, quadratic, and linear algebraic forms applied to graph-theoretic electronic-density and edge-adjacency matrices.
• Matrix Normalization Approaches: Implements double-stochastic and mutual probability matrix normalization formalisms.
• Topological Constraints: Applies topological cut-offs to account for specific inter-atomic relationships.
• Atomic Weighting Schemes: Integrates six atomic properties as weighting schemes in the computation of molecular vectors.
• Local-Fragment Analysis: Considers four local fragments for focused regional descriptor computation.
• Lone-Pair Electron Consideration: Represents lone-pair electron counts via diagonal coefficients in matrix representations.
• Aggregation Operators: Computes global indices by applying several aggregation operators (invariants) over atom- and bond-level descriptors.
• Parallel and Batch Processing: Supports parallel computation of indices and batch processing for large-scale descriptor calculation and data curation.

Scientific Applications:

• Compound Screening: Applied to screening chemical compounds including antimalarials, antibacterials, and tyrosinase inhibitors.
• Diversity Analysis and Descriptor Comparison: Used for diversity analysis and comparative studies with DRAGON descriptors, demonstrated by principal component analysis and Shannon's entropy-based variability analysis.
• QSAR and Binding-Affinity Modeling: Employed in QSAR studies, including corticosteroid-binding globulin binding-affinity analysis using Cramer's steroid dataset.

Methodology:

Implemented in Java v1.7 using the Chemistry Development Kit (CDK) v1.4.19; computations apply graph-theoretic electronic-density and edge-adjacency matrices with bilinear, quadratic, and linear algebraic forms, double-stochastic and mutual probability normalizations, topological cut-offs, six atomic weighting schemes, four local-fragment analyses, lone-pair diagonal coefficients, aggregation invariants, and parallel/batch computation.