Life Science

A Physics-Based Method For Predicting Blood-Brain Barrier Penetration

The human body is designed to protect against outside invaders. Skin acts as a physical barrier to keep viruses and bacteria from entering the bloodstream. Our immune system creates antibodies to fight specific germs.

The blood-brain barrier provides another important defense mechanism—a membrane of cells that filter blood and guard the brain from toxins. Transporter proteins within this membrane recognize foreign substances and keep them outside of the brain. The blood-brain barrier is essential to keeping the brain safe, but it also presents challenges for the development of therapeutics for central nervous system (CNS) diseases, such as Alzheimer’s, multiple sclerosis, and Parkinson’s disease, among others. 

The blood-brain barrier is somewhat permeable, but designing molecules that penetrate the brain successfully is no trivial task. According to a recent analysis, it takes 20 percent longer to develop CNS drugs compared with non-CNS drugs, due in part to the challenge of crossing the blood-brain barrier.

When developing a CNS small molecule drug, one important property medicinal chemists must account for is the unbound brain-to-plasma partition coefficient, also known as Kp,uu. This metric is a ratio that indicates distribution of a drug between plasma and the brain. It provides a direct quantitative measurement of blood-brain barrier penetration and is a key parameter to support decision making for CNS drug discovery teams.

At Schrödinger, our internal therapeutics group has been actively working on several preclinical CNS drug discovery projects. When we started these efforts, we surveyed existing computational tools and realized none of them were accurate at predicting brain Kp,uu—so we set out to create one. 

The knowledge our therapeutics group gains from our R&D efforts is fed back into the platform, advancing its capabilities to benefit our team and all of our customers.

While working on an earlier CNS drug discovery project, where large amounts of brain PK data was available, we observed a strong correlation between brain Kp,uu and E-sol (energy of solvation). E-sol is a fundamental molecular property, calculated with quantum mechanics, that measures the energy required for a compound to go from the gas to water phase. This correlation between Kp,uu and E-sol prompted us to further validate this method with a larger set of molecules and across different projects. Soon enough, we realized that E-sol could be predictive of blood-brain barrier penetration.

We then applied this in silico physics-based method to a curated library of molecules from additional projects including public data sets. We found that E-sol predicts Kp,uu with 79 percent accuracy, which is significantly higher than any existing computational tools. This physics-based method we developed can also be combined with machine learning methods to screen large chemical libraries for potential compounds with favorable brain penetration. 

Application of E-sol has already helped therapeutics group CNS project teams to discover brain penetrant molecules faster and more efficiently. It is also an example of the continuous feedback loop that makes Schrödinger’s computational platform so powerful—the knowledge our internal Therapeutics Group gains from our R&D efforts is fed back into the platform, advancing its capabilities to benefit our team and all of our customers. 

 We have recently implemented the Kp,uu prediction protocol in LiveDesign, Schrödinger’s enterprise informatics platform, and continue to leverage it across multiple internal projects with success. We are also creating a well-defined workflow so external teams can also implement E-sol into their drug discovery projects. We believe that this highly accurate physics-based Kp,uu prediction method could help medicinal chemists and drug discovery teams to design better CNS therapeutics, faster.

Author Photo: Eric Therrien

Eric Therrien

Eric Therrien is a director in the therapeutics group at Schrödinger. Since 2016, Eric has been working as a computational and medicinal chemist leading efforts in collaborative and internal drug discovery programs with a focus in applying free energy perturbation methods. He obtained his Ph.D. in organic chemistry from the Université de Montréal in 2005 and conducted post-doctoral research at McGill University in computational chemistry. Eric contributed to a variety of drug discovery programs including oncology, immunology, and neurology and he is co-author of over 30 publications and patents.

Sign Up Today


Sign up to receive quarterly updates with the latest from Extrapolations.