Prediction of human cytochromeP450inhibition using bio‐selectivity induced deep neural network

Abstract

Since the successful debut of AlphaGo, deep learning (DL) techniques have been applied to almost all areas of data sciences and are achieving remarkable milestones. For example, in predicting cytochrome P450 (CYP450) inhibition, DL and other machine learning techniques applied were proven significantly useful. However, currently, most models are focused on how much they can improve compared to previously published methods by using different methodologies and larger data sets without considering bio‐selectivity. This study provides a multitask classification model for five important isoforms of CYP450. Our model uses the most extensive data set of over 150 000 molecules ever used. To consider the stereoisomerism of bio‐selectivity, new descriptors have been developed. We illustrated the discriminating power of the new descriptors with some examples that demonstrated the improvement in the model's performance using three‐dimensional descriptors, which account for the stereoselectivity of drug–protein/enzymes that two‐dimensional models cannot achieve.

Artificial Intelligence in Drug Metabolism and Excretion Prediction: Recent Advances, Challenges, and Future Perspectives

“…Moreover, especially in recent years, there have been a lot of studies focusing on the prediction of specific metabolic CYP isoforms with remarkably good performance. Many prominent studies have focused on predicting the 5 major CYP inhibitors (1A2, 2C19, 2C9, 2D6, and 3A4), DeepCYP [41], SuperCYPsPred [44], CYPlebrity [43], iCYP-MFE [46], VirtualRat [9], and others [47,48]. Some studies focused on CYP substrate prediction, such as [33,47,49].…”

AiKPro: deep learning model for kinome-wide bioactivity profiling using structure-based sequence alignments and molecular 3D conformer ensemble descriptors

“…One major limitation is that most of these methods do not adequately account for the three-dimensional (3D) structure of proteins and compounds, which is a critical factor influencing their interactions and binding affinities. To address this limitation, more advanced descriptors, such as molecular 3D conformer ensemble descriptors (3CED), have been proposed to better represent the structurally dynamic state of molecules and improve the accuracy of interaction prediction between small molecules and proteins 33 . In addition to 3D conformer ensemble representation of compounds, incorporating protein structure information into kinase inhibitor profiling models is crucial for achieving high predictive performance 31 , 34 – 36 .…”

AiKPro: Deep Learning Model for Kinome-Wide Bioactivity Profiling Using Structure-based Sequence Alignments and Molecular 3D Conformer Ensemble Descriptors

“…One major limitation is that most of these methods do not adequately account for the threedimensional (3D) structure of proteins and compounds, which is a critical factor in uencing their interactions and binding a nities. To address this limitation, more advanced descriptors, such as molecular 3D conformer ensemble descriptors (3CED), have been proposed to better represent the structurally dynamic state of molecules and improve the accuracy of interaction prediction between small molecules and proteins 33 . In addition to 3D conformer ensemble representation of compounds, incorporating protein structure information into kinase inhibitor pro ling models is crucial for achieving high predictive performance 31,[34][35][36] .…”