In CNS drug discovery, the estimation of brain exposure to lead compounds is critical for their optimization. Compounds need to cross the blood–brain barrier (BBB) to reach the pharmacological targets in the CNS. The BBB is a complex system involving passive and active mechanisms of transport and efflux transporters such as P-glycoproteins (P-gp) and breast cancer resistance protein (BCRP), which play an essential role in CNS penetration of small molecules. Several in vivo, in vitro, and in silico methods are available to estimate human brain penetration. Preclinical species are used as in vivo models to understand unbound brain exposure by deriving the Kp,uu parameter and the brain/plasma ratio of exposure corrected with the plasma and brain free fraction. The MDCK-mdr1 (Madin Darby canine kidney cells transfected with the MDR1 gene encoding for the human P-gp) assay is the commonly used in vitro assay to estimate compound permeability and human efflux. The in silico methods to predict brain exposure, such as CNS MPO, CNS BBB scores, and various machine learning models, help save costs and speed up compound discovery and optimization at all stages. These methods enable the screening of virtual compounds, building of a CNS penetrable compounds library, and optimization of lead molecules for CNS penetration. Therefore, it is crucial to understand the reliability and ability of these methods to predict CNS penetration. We review the in silico, in vitro, and in vivo data and their correlation with each other, as well as assess published experimental and computational approaches to predict the BBB penetrability of compounds.

Computational methods for generating molecules with specific physiochemical properties or biolog- ical activity can greatly assist drug discovery efforts. Deep learning generative models constitute a significant step towards that direction. In this work, we introduce a novel approach that utilizes a Reinforcement Learning paradigm, called proximal policy optimization (PPO), for optimizing chemical molecules in the latent space of a pre-trained deep learning generative model. Working in the latent space of a generative model allows us to bypass the need for explicitly defining chemical rules when computationally designing molecules. The generation of molecules with desired properties is achieved through navigating the latent space for identifying regions that correspond to molecules with desired properties. Proximal policy optimization is a state-of-the-art policy gradient algorithm capable of operating in continuous high dimensional spaces in a sample-efficient manner. We have paired our optimization framework with the latent spaces of two autoencoder models, a variational autoencoder and an autoencoder trained with mutual information machine loss respectively, showing that the method is agnostic to the underlying architecture. We present results on commonly used benchmarks for molecule optimization that demonstrate that our method has comparable or even superior performance to state-of-the-art approaches. We additionally show how our method can generate molecules that contain a pre-specified substructure while simultaneously optimizing for molecular properties, a task highly relevant to real drug discovery scenarios.

In CNS drug discovery, estimation of brain exposure of lead compounds is critical for their optimization. Compounds need to cross the blood-brain barrier (BBB) to reach the pharmacological targets in the CNS. The BBB is a complex system involving passive and active mechanisms of transport and efflux transporters such as P-glycoproteins (P-gp) and breast cancer resistance protein (BCRP), which play an essential role in CNS penetration of small molecules. Several in vivo, in vitro and in silico methods are available to estimate human brain penetration. Preclinical species are used as in vivo models to understand the unbound brain exposure by deriving the Kp,uu parameter, the brain/plasma ratio of exposure corrected with the plasma and brain free fraction. MDCK-mdr1 (Madin Darby canine kidney cells transfected with the MDR1 gene encoding for the human P-gp) assay is the most commonly used in vitro assay to estimate compound permeability and human efflux. The in silico methods to predict brain exposure, such as CNS MPO, CNS BBB scores and various Machine Learning models, help save costs and speed up compound discovery and optimization at all stages. These methods enable the screening of virtual compounds and building of a CNS penetrable compounds library as well as optimization of lead molecules for CNS penetration. Therefore, it is crucial to understand these methods' reliability and ability to predict CNS penetration. We review the in silico, in vitro and in vivo data and their correlation with each other and assess published in silico approaches to predict the BBB penetrability of compounds.