Proteomic Learning of Gamma-Aminobutyric Acid (GABA) Receptor-Mediated Anesthesia

TL;DR

This work addresses the need for safer, more controllable anesthetics by leveraging proteomic learning across 24 GABA receptor–related PPI networks. It integrates ML models with NLP-based molecular embeddings to predict binding affinities, assess off-target effects, and perform ADMET screening, culminating in the identification of near-optimal lead compounds and repurposing candidates. The study demonstrates how cross-target BA correlations and hERG risk filtering can guide drug design, with two repurposing-ready leads and several optimized anesthetic analogs backed by docking validations. The approach offers a scalable framework for rapid discovery and optimization of GABA receptor–targeted anesthetics with improved safety profiles and potential for personalized application in anesthesia.

Abstract

Anesthetics are crucial in surgical procedures and therapeutic interventions, but they come with side effects and varying levels of effectiveness, calling for novel anesthetic agents that offer more precise and controllable effects. Targeting Gamma-aminobutyric acid (GABA) receptors, the primary inhibitory receptors in the central nervous system, could enhance their inhibitory action, potentially reducing side effects while improving the potency of anesthetics. In this study, we introduce a proteomic learning of GABA receptor-mediated anesthesia based on 24 GABA receptor subtypes by considering over 4000 proteins in protein-protein interaction (PPI) networks and over 1.5 millions known binding compounds. We develop a corresponding drug-target interaction network to identify potential lead compounds for novel anesthetic design. To ensure robust proteomic learning predictions, we curated a dataset comprising 136 targets from a pool of 980 targets within the PPI networks. We employed three machine learning algorithms, integrating advanced natural language processing (NLP) models such as pretrained transformer and autoencoder embeddings. Through a comprehensive screening process, we evaluated the side effects and repurposing potential of over 180,000 drug candidates targeting the GABRA5 receptor. Additionally, we assessed the ADMET (absorption, distribution, metabolism, excretion, and toxicity) properties of these candidates to identify those with near-optimal characteristics. This approach also involved optimizing the structures of existing anesthetics. Our work presents an innovative strategy for the development of new anesthetic drugs, optimization of anesthetic use, and deeper understanding of potential anesthesia-related side effects.

Figures (6)

• Figure 1: Flowchart of nearly optimal lead compounds screening for Gamma-aminobutyric acid (GABA) receptor agonists. a: The protein-protein interaction (PPI) networks of 24 GABA receptor subtypes involve 4824 proteins, and each receptor subtype has a core and global PPI network. Here only two PPI networks (GABRA1 and GABRA5) with several compounds are shown for simplicity. For more detailed information on the PPI networks, please refer to Table of the Supporting Information. b: The drug target interaction (DTI) network constructed against GABA receptors include 136 targets and 183250 inhibitor compounds that are collected from ChEMBL database in c. Here, only four targets (GABRA1, GABRA2, GABRA5, and GABBR2) with a few compounds are presented for simplicity. The yellow dashed lines mean the connections among 136 targets. d: Nearly optimal lead compounds were screened by two technical routes, the first being a predictive model for side effects and repurposing assessment as well as an ADMET screening model, and the second being molecular optimization of existing drugs.
• Figure 2: Examples of side effect and repurposing potential prediction. a: Three rows of inhibitors targeting GABRA5 are presented, each causing side effects on zero, one, and both of the two side effect targets, respectively. The yellow frames indicate no side effects. b: Inhibitors of GABRA5 with repurposing potential are shown, where the pink frames indicate an inhibitor's repurposing potential for one therapeutic target without side effects on the other.
• Figure 3: Six examples of related predicted BAs illustrate the sequence and structural similarities of proteins. In each example, the $x$ and $y$ axis of the panel display the predicted BA values for two other proteins. On the right side of the scatter plot, the 3D structural alignment is shown, while the 2D sequence alignment is displayed below. The 3D structures used for alignment include PDB 7BU7 and 4LDE for ADRB1 and ADRB2 (a), PDB 7VIE and 7T6B for S1PR1 and S1PR2 (b), PDB 6ZFZ, 3UON, and 5DSG for CHRM1, CHRM2, and CHRM4 (c), PDB 7MIX and 8WE6 for CACNA1B and CACNA1C (d), PDB 5U09 and 5ZTY for CNR1 and CNR2 (e), and PDB 7CX2 and 7WU9 for PTGER2 and PTGER3 (f).
• Figure 4: ADMET properties, SAS, and hERG side effects screening for the datasets of GABRA5, SCN9A, CNR1, SLC6A3, and SLC6A5. The color of the scatter points represents the experimental BA values of the compounds in each dataset, with green frames highlighting the optimal range for properties and side effects.
• Figure 5: Further evaluation of additional ADMET properties for the identified repurposable molecular compounds is conducted. a and c represent the predicted ADMET properties, chemical graph and side effect assessments for compound ChEMBL1372447, while b and d represent these predictions and graphs for compound ChEMBL200482. In a and b, the boundaries of the yellow and orange regions respectively highlight the upper and lower limits of the optimal range for ADMET properties. The blue curves indicate the values of the specified 13 ADMET properties. The predictions shown in a and b are from the ADMETlab 2.0 website (https://admetmesh.scbdd.com/). The 3D docking structures between compunds ChEMBL1372447, ChEMBL200482, and GABRA5 are shown in e and g. The corresponding 2D interaction diagrams are given in f and h. The PDB ID for the GABRA5 protein is 8BHG. AutoDock Vina was used for protein-ligand docking, and hydrogen bonds play a crucial role in binding energy. Abbreviations: MW (Molecular Weight), $\log$P (log of octanol/water partition coefficient), $\log$S (log of the aqueous solubility), $\log$D (logP at physiological pH 7.4), nHA (Number of hydrogen bond acceptors), nHD (Number of hydrogen bond donors), TPSA (Topological polar surface area), nRot (Number of rotatable bonds), nRing (Number of rings), MaxRing (Number of atoms in the biggest ring), nHet (Number of heteroatoms), fChar (Formal charge), and nRig (Number of rigid bonds).