U.S. flag An official website of the United States government

On Oct. 1, 2024, the FDA began implementing a reorganization impacting many parts of the agency. We are in the process of updating FDA.gov content to reflect these changes.

  1. Home
  2. Drugs
  3. Science and Research | Drugs
  4. Regulatory Science in Action
  5. New developments in regulatory QSAR modeling: a new QSAR model for predicting blood brain barrier permeability
  1. Regulatory Science in Action

New developments in regulatory QSAR modeling: a new QSAR model for predicting blood brain barrier permeability

Quantitative structure-activity relationship (QSAR) models are becoming an increasingly important part of regulatory review because they can provide rapid assessment of the toxicological and pharmacological properties of a compound based solely on its chemical structure. QSAR modeling is a well-known technique that reveals associations between structural characteristics or properties and biological or toxicological activities under the general assumption that similar chemical structures display similar activities. Predictions can be generated for a drug substance itself, intermediates, pre-cursor materials related to the drug substance, degradation products, or leachables. Additionally, predictions can be used to prioritize experimental inquiry into the potential effects of newly identified drugs of abuse and assist with emergency scheduling. QSAR models provide regulatory utility in cases where empirical data are limited or unavailable, and an assessment of toxicological/pharmacological potential at endpoints of regulatory interest is needed.

Glossary

Coverage: the percent of chemicals for which a prediction can be made

cross validation: a procedure in which random portions of the data set used to train the model are withheld and the model is refitted using the remaining data. Repeating this process and averaging the performance statistics generally results in a better estimate of how the model will perform when used on new data.

feature: a characteristic of a molecule (e.g., a property, characteristic, or structural component of a molecule)

negative predictive value: the percentage of negative predictions that are true negatives

sensitivity: the percentage of known positives that are correctly predicted

specificity: the percentage of known negatives that are correctly predicted

QSAR models: mathematical models that describe the correlation between specific characteristics, or structural features, of molecules and their chemical or biological activities (for example tendency to cross the blood-brain barrier) under the general assumption that similar structures give rise to similar activities).

The Division of Applied Regulatory Science’s (DARS) computational toxicology and pharmacology research is focused on the development of highly curated data sets and QSAR models for endpoints of regulatory interest. In a recent effort, two statistical-based QSAR models were developed to predict drug permeability across the blood brain barrier (BBB). The intended purpose of these models is to assist regulators with abuse liability assessment of drug metabolites or other materials related to drug substances. Furthermore, these models can be used to predict whether an unknown substance of abuse can permeate the BBB to produce effects predicted by the Public Health Assessment via Structural Evaluation (PHASE) approach. PHASE is a multi-component computational approach that determines if newly emerging drug substances on the street-drug market are a risk to public safety. This approach has been previously applied in the evaluation of fentanyl derivatives and kratom alkaloids. Lastly, BBB models can be utilized by pharmaceutical companies to identify novel therapeutics that target central nervous system (CNS) disorders.

To develop reliable QSAR models that would be applicable to a wide range of chemicals including pharmaceuticals and substances of abuse, DARS researchers harvested data for approximately one thousand molecules from multiple publicly available data sources. The models were trained exclusively on results from in vivo experiments in rodents in which the ratio of drug in the brain to that in blood or plasma was determined. These data are considered the gold standard for determining the extent to which a drug can cross the BBB. Two different commercial software packages were then used to construct the models and classification thresholds (i.e., the minimum ratio at which a drug is designated as permeable rather than impermeable) were chosen for both models based on the tradeoff between the specificity and sensitivity of their predictions.

QSAR Figure 1 - QSAR Algorithm, Experimental activity data, Chemicals Structures pointing to QSAR Model.  Unknown Chemical and Activity Prediction are shown next to QSAR Model

Figure 1. QSAR models for BBB permeability were developed using two different computer algorithms that identified and extracted relationships between a set of chemical structures and their biological activities. Newly developed QSAR models can be used to predict if an unknown chemical is capable of crossing the BBB.

Predictive performance of the models was assessed using both cross-validation and external validation. The cross-validation performance ranged from 82-85% in sensitivity and 80-83% in negative predictivity. The external validation was performed using a set of 83 chemicals (42 BBB permeable and 41 BBB impermeable) obtained from published literature. Overall, performance of individual models ranged from 70-75% in sensitivity, 70-72% in negative predictive value and 78-86% in coverage. The predictive performance was further improved to 80% in sensitivity and 93% in coverage by combining predictions across the two software programs. The improved performance indicates that the two models are complementary and greater confidence can be inferred when the predictions are in agreement amongst both software.

Models and databases developed through the DARS are used internally to inform and support regulatory decisions for drug products. Additionally, the models are made available to the public through commercial QSAR software vendors with whom CDER has formal Research Collaboration Agreements. These new models can be used to rapidly predict whether a molecule is getting into the brain and reduce the need for in vivo testing. Identification of drug candidates or other substances in the finished drug product that cross the BBB can inform strategies for reducing the potential for abuse liability in a regulatory setting and can assist drug developers with designing CNS drugs.

How does this work advance drug development?

Identification of drug candidates or other substances in the finished drug product that cross the BBB can inform strategies for reducing the potential for abuse liability in a regulatory setting and can assist drug developers with designing drugs intended to act on targets in the central nervous system.

Reference

Faramarzi, Sadegh, Marlene T. Kim, Donna A. Volpe, Kevin P. Cross, Suman Chakravarti, and Lidiya Stavitskaya. "Development of QSAR models to predict blood-brain barrier permeability." Frontiers in Pharmacology (2022): 4486.


Back to Top