TL;DR
Researchers have developed a new, simple metric called Normalized Bond Energy (NBE) to predict how "druggable" small molecules are. This metric shows promise in quickly assessing key properties like membrane permeability and blood-brain barrier penetration, potentially streamlining early-stage drug discovery.
What Problem Does This Paper Address?
Drug discovery is a lengthy and expensive process, with a significant portion of potential drug candidates failing. Chemical small molecules (CSMs) are a major class of drugs, but many have low "druggable potential." Predicting this potential early on is crucial. While existing in-silico methods can estimate properties like logP and pKa, they often fall short, especially for diverse chemical structures, leading to high-risk screening. There's a clear need for new, efficient, and reliable metrics to quantify the druggability of CSMs.
Methods
The core innovation here is the Normalized Bond Energy (NBE) metric. The process involves:
- Extracting Bonds: Using the RDKit library, the algorithm identifies all chemical bonds within a CSM.
- Determining Bond Energy: Each bond's energy is calculated by matching its type (single, double, triple) and the types of the atoms it connects to a pre-defined bond energy table (provided in Supplement Table 1). This energy quantifies the strength of a chemical bond, representing the energy required to break it.
- Calculating Total Bond Energy: The energies of all bonds in the molecule are summed up.
- Normalization: To account for the fact that larger molecules naturally have more bonds and thus higher total bond energy, the total bond energy is divided by the molecule's molecular weight. This normalization step is key to making the metric comparable across molecules of different sizes.
The paper also details statistical analyses performed, including Spearman's correlation analysis, t-tests, and Wilcoxon tests, to evaluate the relationships between NBE and various druggable properties.
Datasets
The study utilized two primary datasets:
- DrugBank Database: This dataset contains information on approved and experimental small molecule drugs.
- Human Metabolome Database (HMDB): This dataset includes human endogenous metabolites.
Additionally, curated literature data was used for specific druggable properties such as Caco-2 monolayers permeability, blood-brain barrier penetration, and human intestinal absorption.
Key Insights and Ideas
- Bond Energy as a Druggability Proxy: The fundamental idea that the "energy status" of a molecule's bonds can reflect its druggable properties is a novel perspective. This moves beyond traditional descriptors that focus solely on molecular structure or physicochemical properties.
- Normalization is Crucial: Recognizing that larger molecules have more bonds and thus higher raw bond energy is critical. Normalizing by molecular weight allows for a more equitable comparison of NBE across diverse CSMs.
- Metabolites as a Druggability Benchmark: Applying NBE to human endogenous metabolites, which are naturally occurring CSMs, provides a unique angle. It suggests that NBE can also characterize biological molecules and their distribution within the body.
- NBE Reflects Multiple Druggable Properties: The finding that NBE correlates with a range of important properties (logP, pKa, permeability, BBB penetration, intestinal absorption) suggests it's a versatile metric, not just a proxy for one specific characteristic.
Key Results
- Approved Drugs vs. Unapproved: Approved drugs in the DrugBank dataset showed significantly higher NBE scores than unapproved drugs.
- Solubility and NBE: Soluble CSMs had smaller NBE scores compared to insoluble ones.
- Correlations with Physicochemical Properties: NBE showed significant correlations with:
- Melting Point: Negative correlation (Rho = -0.19).
- logP: Positive correlation (Rho = 0.36).
- pKa: Positive correlation (Rho = 0.38).
- Correlations with Transport Properties: NBE demonstrated strong correlations with:
- Caco-2 Monolayers Permeability: High correlation coefficients (e.g., Rho = 0.80).
- Human Intestinal Absorption: Significant positive correlations (e.g., Rho = 0.44 for HIA+ vs. HIA-).
- Blood-Brain Barrier Penetration: Significant positive correlations (e.g., Rho = 0.55 for BBB+ vs. BBB-).
- Metabolite Analysis: NBE scores differed significantly across various body fluids and cellular locations, and correlated with melting point and water solubility in the HMDB dataset, mirroring findings from CSMs in DrugBank.
Limitations and Future Work
- Limitations: The paper acknowledges that while NBE is a promising metric, it's still an in-silico method and its predictive power might be limited by the complexity of biological systems. The authors also note that the bond energy table used is a general one, and specific bond environments might influence actual bond energies.
- Future Work: The authors suggest that NBE could be further improved and validated by combining it with other in-silico methods or metrics. They also propose exploring its application in more complex biological contexts and for predicting other druggable properties. The potential for NBE to provide valuable clues for metabolite-based drug discovery is also highlighted as a future avenue.
Summary updated