Publication: INTEGRATION OF ARTIFICIAL INTELLIGENCE (AI) AND COMPUTER AIDED DRUG DISCOVERY (CADD) TECHNIQUES FOR THE DEVELOPMENT OF NOVEL DRUG-LIKE KEAP1-NRF2 INHIBITORS/DIRECT NRF2 ACTIVATORS
| dc.contributor.author | MAK KIT KAY | |
| dc.date.accessioned | 2024-03-10T00:35:58Z | |
| dc.date.available | 2024-03-10T00:35:58Z | |
| dc.date.issued | 2023 | |
| dc.description.abstract | Nuclear factor (erythroid-derived 2)-like 2 (Nrf2) is a master transcriptional factor that regulates antioxidative responses during inflammation. Recent scientific evidence has suggested that therapeutic targeting of Nrf2 would pave the way for discovering new therapeutic interventions for treating neuroinflammatory diseases such as Alzheimer’s disease, Parkinson's disease, multiple sclerosis, etc. Very recently, US-FDA has approved two Nrf2 activators (dimethyl fumarate and monomethyl fumarate) for the treatment of patients with relapsing forms of multiple sclerosis. However, these two drugs are electrophilic activators that elicit their action by reacting with cysteine residues of Kelch-like ECH-associated protein 1 (Keap1), which regulates the Nrf2 activity and acts as a sensor for oxidative and electrophilic stresses. However, these two US-FDA-approved Nrf2 activators do produce adverse events as they can bind the cysteine residues of other proteins in the biological system. Thus, a hypothesis of developing non-electrophilic Nrf2 activators was put forward for developing therapeutic interventions for neuroinflammatory diseases. In this study, I have attempted to identify the novel templates of non-electrophilic Nrf2 activators by integrating artificial intelligence (AI) and computer-aided drug discovery (CADD) techniques. The Konstanz Information Miner (KNIME), an open-source data analytics platform, was used to construct the workflows. The workflows contain 1) data analysis, 2) machine learning, 3) cheminformatics (RD kit and ChemAxon), and 4) Schrodinger discovery suite nodes. The machine learning models were trained based on the bioactivity (Nrf2) data in the ChemBL database. Two chemical databases, 1) an in-house database of secondary metabolites in Malaysian medicinal plants and 2) Enamine’s CNS library of compounds, were fed into the workflows for identifying the new chemical templates of non-electrophilic Nrf2 activators. The test compounds used in this study are 1) three natural products (Swietenine, Auranamide, and Patriscabratine), 2) twenty-three new 6-Shogaol inspired synthetic compounds, and 3) fifty-seven new sulforaphane-inspired synthetic compounds, 2-amino-4,5,6,7-tetrahydrobenzo[b]thiophenes (THBTs). Swietenine was isolated from Swietenia macrophylla seeds using conventional and super-critical fluid extraction methods. Auranamide and Patriscabratine were isolated from Melastoma malabathricum leaves. They were also synthesised using L-phenylalanine as a starting material. The cell lines used in this study were 1) murine hepatoma cells (Hepa-1c1c7), 2) murine macrophages (RAW264.7), and 3) murine microglia (BV-2). The non-toxic dose of the test compounds on the cell lines was assessed using MTT (3-(4, 5- dimethylthiazolyl-2)-2, 5-diphenyltetrazolium bromide) assay, DAPI (4′,6-diamidino- 2-phenylindole) staining, and observation under the microscope. The Nrf2 activity of the test compounds was assessed using a prochaksa microtiter plate assay using Hepa- 1c1c7 cells and further confirmed by determining the genes and proteins expression associated with Nrf2 activation (Nrf2, heme oxygenase-1 (HO-1) and NQO1). The anti-inflammatory activity of the test compounds was assessed using the Griess assay in Nrf2-proficient and Nrf2-deficient RAW264.7 and BV-2 cells. The mechanisms involved in the anti-inflammatory activity of the test compounds were assessed by determining the genes and proteins expression associated with inflammatory mediators (IL-1, IL-6, TNF-, IFN-, COX-2, and NF-B). The expression of genes was quantified using RT-qPCR (Reverse transcription quantitative real-time polymerase chain reaction (PCR)) and following 2-Ct method. The protein expression was quantified using commercially available ELISA (enzyme-linked immunosorbent assay) kits. The Nrf2 activity of swietenine was further elaborated by testing its bioactivity in streptozotocin/high-fat diet-induced non-alcoholic fatty liver disease (NAFLD) in diabetic mice. The biochemical markers of NAFLD in the blood (glucose, cholesterol, triglycerides, alanine transaminase (ALT), aspartate transaminase (AST), alkaline phosphatase (ALP), glutathione (GSH), total antioxidant capacity (TAC), and malondialdehyde (MDA)) were assessed using the commercially available kits. The cholesterol and triglyceride levels in the liver were also measured. The liver index and hepatic lipid accumulation (using oil-O-red staining) were assessed. Gene expression (RT-qPCR) and immunohistochemical studies were performed to elucidate the mechanisms by determining the genes and proteins expression associated with lipogenesis (ACLY, ACC1, FASN, SREBP1c, and ChREBP) and oxidative stress (Nrf2, NQO-1 and HO-1). The metabolic stability of all the test compounds was assessed in mouse, rat and human liver microsomes by measuring the intrinsic clearance (Clint), elimination rate constant (Kel) and calculating half-life (T1/2). The concentration of the test compounds was measured at different time intervals to calculate Clint, Kel and T1/2. The insights on the molecular interactions of the test compounds with amino acid residues of the Keap1’s kelch domain were obtained by performing molecular docking and molecular dynamics studies using Schrodinger software. The binding energies of the test compounds in the binding pockets of the kelch domain were calculated using Schrodinger software. The drug-like properties of the test compounds were also predicted using Schrodinger software. The compounds predicted from the KNIME workflow, in which machine learning nodes were integrated with molecular docking nodes, have activated Nrf2 in the micromolar range by inhibiting its interactions with the kelch domain of Keap1. All the test compounds have shown Nrf-2-dependent anti-inflammatory activity and downregulated the levels of pro-inflammatory cytokines. All the test compounds are drug-like and formed stable interactions with the critical amino acids in the kelch domain of Keap 1. The KNIME workflow developed in this work can be used to identify the new drug-like compounds against any given biological target of interest. The concept of designing natural product-inspired (6-Shogaol and Sulforaphane) drug-like molecules adopted in this study can be extended to other natural products containing flexible side chains and reactive electrophilic pharmacophoric features. The reusable catalyst, proline-proline dipeptide, developed in this study can be used in other types of reactions. This study has provided new chemical templates that can be optimised to produce safe and effective Nrf2 activators with potential applications in treating neurodegenerative diseases. | |
| dc.identifier.uri | https://hdl.handle.net/20.500.14377/36074 | |
| dc.language.iso | en | |
| dc.publisher | International Medical University | |
| dc.subject | Artificial Intelligence | |
| dc.subject | Inflammation | |
| dc.subject | Neuroinflammatory Diseases | |
| dc.subject | Drug Discovery | |
| dc.subject | Immune System | |
| dc.subject | Machine Learning | |
| dc.subject | Neurodegenerative Diseases | |
| dc.title | INTEGRATION OF ARTIFICIAL INTELLIGENCE (AI) AND COMPUTER AIDED DRUG DISCOVERY (CADD) TECHNIQUES FOR THE DEVELOPMENT OF NOVEL DRUG-LIKE KEAP1-NRF2 INHIBITORS/DIRECT NRF2 ACTIVATORS | |
| dc.type | Thesis | |
| dspace.entity.type | Publication | |
| oairecerif.author.affiliation | #PLACEHOLDER_PARENT_METADATA_VALUE# |