A deep learning approach to predict blood-brain barrier permeability

Abstract

The blood–brain barrier plays a crucial role in regulating the passage of 98% of the compounds that enter the central nervous system (CNS). Compounds with high permeability must be identified to enable the synthesis of brain medications for the treatment of various brain diseases, such as Parkinson’s, Alzheimer’s, and brain tumors. Throughout the years, several models have been developed to solve this problem and have achieved acceptable accuracy scores in predicting compounds that penetrate the blood–brain barrier. However, predicting compounds with “low” permeability has been a challenging task. In this study, we present a deep learning (DL) classification model to predict blood–brain barrier permeability. The proposed model addresses the fundamental issues presented in former models: high dimensionality, class imbalances, and low specificity scores. We address these issues to enhance the high-dimensional, imbalanced dataset before developing the classification model: the imbalanced dataset is addressed using oversampling techniques and the high dimensionality using a non-linear dimensionality reduction technique known as kernel principal component analysis (KPCA). This technique transforms the high-dimensional dataset into a low-dimensional Euclidean space while retaining invaluable information. For the classification task, we developed an enhanced feed-forward deep learning model and a convolutional neural network model. In terms of specificity scores (i.e., predicting compounds with low permeability), the results obtained by the enhanced feed-forward deep learning model outperformed those obtained by other models in the literature that were developed using the same technique. In addition, the proposed convolutional neural network model surpassed models used in other studies in multiple accuracy measures, including overall accuracy and specificity. The proposed approach solves the problem inevitably faced with obtaining low specificity resulting in high false positive rate.

Design, pharmacokinetic profiling, and assessment of kinetic and thermodynamic stability of novel anti-Salmonella typhi imidazole analogues

“…Ligands R, B-1, B-2, with estimated solubility (ESOL) value of 15.7 mg/ml, 0.006 mg/ml, and 0.009 mg/ml, respectively, were found to exhibit moderate to high solubility (Table 4). Likewise, Blood Brain Barrier (BBB) is an endothelial cell layer of the brain that restrict entry of toxins and xenobiotic into the brain (Alsenan et al 2021;Tong et al 2022;Wang et al 2019). The result in Table 4 shows that all the ligands possess no BBB permeating potentials.…”

Drug Repurposing, a Fast-Track Approach to Develop Effective Treatments for Glioblastoma

“…This approach requires, already from a preclinical stage, a focus on drugs showing a higher probability of crossing the BBB. Various algorithms and drug discovery tools have been developed to predict the CNS permeability of drugs [ 178 , 179 , 180 , 181 , 182 , 183 , 184 ]. Among these is the well-recognized CNS Multiparametric Optimization (CNS-MPO) desirability tool, which is characterized by a simple-to-use design algorithm and multiparameter approach in drug discovery [ 40 ].…”

Predicting In Vivo Compound Brain Penetration Using Multi-task Graph Neural Networks

“…Since most publicly available brain penetration data is categorical, literature is directed toward classification models that discriminate between brain-penetrant (BP+) and non-brain-penetrant (BP−) compounds. For such classification, a K p value of 0.1 is commonly used as the threshold to classify compounds as brain-penetrant ( K p > 0.1) or non-penetrant ( K p ≤ 0.1). ,− This definition typically leads to imbalanced data sets dominated by BP+ compounds, − ,, which imposes further modeling challenges and requires the use of adequate performance metrics, such as Matthew’s correlation coefficient (MCC) . Regression models were less common and usually based on reduced data sets. ,,, …”

“…So far, publicly available data was mainly used for the development and evaluation of brain penetration models. − ,, Despite some valuable efforts toward data set aggregation and standardization, brain penetration data sets still remain heterogeneous and comparatively small for ML approaches, ranging from few hundreds to few thousands of molecules. − ,,, The most recently published data set “B3DB” constitutes the largest publicly available in vivo brain penetration data set and, to the best of our knowledge, it has not been used for modeling yet. Perhaps, due to the limited data set size, previous studies have mainly reported on model performance on random compound subsets, ,− , which is an indicator of self-consistency but not of future model predictivity. , For a more realistic estimation of model prospective performance, evaluation should be done on new chemical series or scaffolds (series or scaffold split) , or on the most recent experiments (temporal split). The latter resembles the model use in pharmaceutical research and requires temporal or date information, which is typically only available in proprietary data sets. , …”