Yeni Mikrozomal Prostaglandin E2 Tip-1 İnhibitörlerinin Yapay Zekâ Destekli Sanal Tarama Çalışmalarıyla Belirlenmesi

Olğaç A. (Executive), Banoğlu E.

Research Project of the Presidency of Turkey Health Institutes (TÜSEB), 2022 - 2023

  • Project Type: Research Project of the Presidency of Turkey Health Institutes (TÜSEB)
  • Begin Date: October 2022
  • End Date: October 2023

Project Abstract

Prostaglandins (PG) are lipid mediators which have many physiological and pathophysiological roles. Their levels increase when inflammation occurs and they intermediate in pain, fever and inflammatory response. The pathway starts with the metabolism of arachidonic acid (AA) to prostaglandin (PG)G2 and PGH2 by cyclooxygenase (COX)-1 and -2, later proceeds with PGE2 synthases (PGES). Cytosolic PGES (cPGES), microsomal PGES Type -1, and -2 (mPGES-1 and -2) convert PGH2 to PGE2. Nonsteroidal anti-inflammatory drugs (NSAID) are frequently used for the treatment of inflammation and inflammation-related diseases. These drugs show their effect by inhibiting COX and suppressing PGE2 production together with other lipid mediators (PGI2, PGD2, and thromboxane (Tx)A2) which also cause reducing essential products for the organism except the disease factors. Especially their long-term usage may cause gastrointestinal bleeding and various cardiovascular complications. Therefore, blocking PGE2 production at a lower step of the pathway via mPGES-1 inhibition is expected to result in a safer and more promising treatment of inflammation, cancer, and cardiovascular diseases. Currently, there is no marketed mPGES-1 inhibitor, but there are ongoing efforts to discover novel compounds and evaluate them in preclinical and clinical trials. The enzyme is formed in a homotrimer structure and embedded in the membrane. Its chains may have either open or closed conformations. The inhibitor binding site is located next to the glutathione (GSH) binding site. The active site is large and surrounded by hydrophobic or flexible polar amino acids and also facing with solvent and membrane. mPGES-1 inhibitors have three different binding patterns; we name them as I, I-II, and I-III binding. Cross-docking studies and other computational analyses on the binding patterns of the cocrystallized mPGES-1 inhibitors showed us the difficulty to catch the correct binding by in silico approaches. Determination of wrong binding modes is also seen in many poses within published papers. This issue is considered to be related to (i) conformational changes occurring with the active site residues, (ii) non-existence of membrane structure (iii) water molecules taking a role in the interaction network and also diverse locations observed for different molecules. Docking scores are obtained from the binding poses, placing the molecule or its fragments into a wrong subpocket limit the success rate of structure-based virtual screening studies. Therefore, we manually curated bioactivity data of 8k mPGES-1 inhibitors to develop reliable binding determination workflows and to train artificial intelligence (AI) models. In our department, by utilizing computer-aided drug design studies, we discovered novel and chemically different scaffolds, then applied various chemical modifications to shed light into their structure-activity relationships. The AI models that we are going to train in this project with convolutional neural networks will support the determination of correct mPGES-1-ligand binding modes and will also allow mPGES-1 specific scoring with better successful prediction rates. In this project we will screen a chemical compound database that consists of approximately 8 million in-stock unique compounds obtained from more than 70 chemical compound vendors. In vitro biological screening studies of identified compounds will be conducted at Friedrich-Schiller University Jena (Germany).