TL;DR
This paper introduces "Normalized Bond Energy" (NBE), a novel in silico metric that quantifies the chemical and biological properties of small molecules. NBE effectively predicts key drug-like properties such as permeability, absorption, and ionization, offering a simpler and more efficient way to screen potential drug candidates and understand metabolite behavior.
What Problem Does This Paper Address?
Developing new drugs is a costly and high-risk endeavor, largely because many candidate small molecules have poor "druggable potential." This means they don't possess the right characteristics to effectively interact with biological targets and reach their intended sites of action within the body. While several computational methods exist to predict these crucial drug-like properties (like logP for lipophilicity or pKa for ionization), identifying truly promising candidates remains challenging. The field needs new, quantitative metrics that can efficiently assess these properties early in the drug discovery pipeline, especially given the vast chemical space of potential drug molecules and the growing interest in using endogenous metabolites as drug leads.
Methods
The core innovation of this paper is the development of Normalized Bond Energy (NBE). The process involves:
- Structure Preparation: Reading the chemical structure of a small molecule (from SDF/SMILES files), adding hydrogen atoms, and converting aromatic bond types to non-aromatic ones.
- Bond Energy Calculation: Matching each bond in the molecule to a predefined "Bond Energy Table" based on bond type (single, double, triple) and atom types. Bond energy itself is defined as the energy required to break a bond.
- Normalization: Summing the energies of all bonds in the molecule and then dividing this sum by the molecule's molecular weight. This normalization step is crucial, as larger molecules naturally have more bonds and thus higher total bond energy.
The authors then used this NBE metric to analyze datasets of drugs and metabolites, correlating NBE scores with experimentally determined properties. Statistical analyses like Spearman correlation, t-tests, and Wilcoxon tests were employed to assess these relationships.
Datasets
The study utilized two primary datasets:
- DrugBank (Version 5.0): This database provided structural data (SDF format) for 10,426 chemical small molecules (CSMs), including 2,444 approved drugs and 7,982 experimental drugs. Experimentally derived properties like melting point, logP, and pKa were also curated from DrugBank.
- Human Metabolome Database (HMDB): This database provided structural data for 113,878 human metabolic CSMs. Experimentally derived properties for these metabolites were also collected.
Additionally, specific datasets were used for evaluating NBE's correlation with transport properties:
- Caco-2 monolayer permeability data: From van De Waterbeemd et al. (1996) and related studies, for 690 CSMs.
- Blood-brain barrier (BBB) penetration data: From Kelder et al. (1999) and related studies, for 1,638 CSMs.
- Human intestinal absorption (HIA) data: From Palm et al. (1997) and Shen et al. (2010), for 598 CSMs.
Key Insights and Ideas
- NBE as a Universal Descriptor: The paper proposes NBE not just for drugs but also for endogenous metabolites, suggesting its potential as a broad descriptor of small molecule properties.
- Solubility and NBE: Soluble CSMs tend to have lower NBE scores than insoluble ones, implying that molecules with weaker overall bond strengths (relative to their size) are more soluble.
- NBE and Physicochemical Properties: NBE shows significant correlations with fundamental properties like melting point (negative correlation), logP (positive correlation), and pKa (positive correlation). This suggests NBE captures aspects of molecular stability, hydrophobicity, and ionization state.
- NBE and Transport Properties: Crucially, NBE strongly correlates with key drug transport properties like Caco-2 permeability, human intestinal absorption (HIA), and blood-brain barrier (BBB) penetration. This is a major finding, as these properties are critical for a drug's efficacy and safety.
- Metabolite Differentiation: NBE scores differ significantly between metabolites found in various body fluids and cellular locations, hinting at its utility in understanding metabolic compartmentalization and function.
Key Results
- Approved vs. Unapproved Drugs: Approved drugs in the DrugBank dataset generally have higher NBE scores than unapproved drugs, suggesting a link between NBE and established druggability.
- Solubility: Soluble CSMs had significantly lower NBE scores (mean 53.3) compared to insoluble CSMs (mean 56.7) (p=0.009).
- Melting Point: NBE showed a significant negative correlation with melting point (Rho = -0.19, p < 1.89e-13).
- LogP: NBE showed a significant positive correlation with logP (Rho = 0.36, p < 1.25e-47).
- pKa: NBE showed a significant positive correlation with pKa (Rho = 0.38, p < 9.08e-19).
- Permeability: NBE demonstrated strong positive correlations with Caco-2 monolayer permeability (e.g., Rho = 0.80, p = 0.01 for one dataset).
- HIA: NBE correlated positively with HIA (e.g., Rho = 0.44, p = 0.05). HIA+ CSMs had higher NBE scores than HIA- CSMs (mean 57.0 vs. 52.4, p=0.0064).
- BBB Penetration: NBE correlated positively with BBB penetration (e.g., Rho = 0.55, p < 9.39e-05). BBB+ CSMs had higher NBE scores than BBB- CSMs (mean 58.7 vs. 54.2, p < 1.08e-15).
- Metabolites: NBE scores showed significant differences between metabolites from different body fluids and cellular locations. NBE was also correlated with melting point (Rho = -0.29, p < 7.11e-138) and water solubility (Rho = -0.29, p < 1.11e-51) in the HMDB dataset.
Limitations and Future Work
- Threshold Limitations: The authors acknowledge that NBE is not a metric that can be used to judge druggable potential solely by a fixed threshold. Instead, it's best used for comparative analysis among a set of molecules.
- Data Source Limitations: The study notes that the number of CSMs in some datasets was small, which might introduce bias and affect the accuracy of the results.
- Future Directions: The authors suggest that NBE's utility could be improved by combining it with other in silico methods or metrics. They also highlight the potential of NBE to provide valuable clues for metabolite-based drug discovery and understanding metabolic compartmentalization. Further research could explore its application in predicting other drug-related properties and in larger, more diverse datasets.
Summary updated