ligands_10k_supplementary data for<b> Information Force Framework: A Conceptual Model for Ligand Ranking in Drug Design</b>

This paper presents a proof-of-concept framework for rapid, interpretable ligand ranking based on information vector theory and ordinary differential equations (ODEs). The method represents molecular properties (pIC50, selectivity, LogP, molecular weight, TPSA, and metabolic clearance) as vectors in semantic space. Ligand candidates are evaluated using weighted-ratio-based normalization toward target properties, generating a force that drives the ODE dynamics describing protein-ligand binding. The resulting Vector Convergence Factor (VCF) and the weighted distance metric provide a transparent, actionable ranking with computational complexity $O(n)$. Empirical validation in a high-throughput data set of 10,000 synthetic ligands produces a strong global correlation ($r \approx 0.99$, $p < 0.001$) between the relative deviation based on the ratio and the distance calculated by the model. Crucially, further analysis reveals that this correlation decreases significantly ($r \approx 0.87$) for the top 5\% of candidates, indicating that the ODE-based model offers distinct, nonlinear discrimination capabilities in the critical lead-optimization zone. The approach demonstrates full interpretability and exceptional scalability, processing 10,000 ligands in approximately 5 milliseconds.