Scholarlink

<p><i>Mycobacterium tuberculosis</i> (Mtb) is a rod-shaped, non-motile, and entirely aerobic bacterium, which is responsible for Tuberculosis. Mtb’s remarkable ability to adapt and survive in hostile conditions makes it one of the most resilient and drug-resistant pathogens. One of the key systems that supports this adaptability is the Two-Component Regulatory System (TCRS), which helps Mtb respond to various environmental stresses. In this study, we explored the GSE6750, GSE182749, and GSE53640 microarray datasets of Mtb, specifically focusing on the deletion mutant mprAB, regX3, and MtrA TCRS. Initially, we characterized the primary transcriptome data, resulting in 125 dysregulated genes (DGs) for mprAB, 388 DGs for regX3, and 250 DGs for MtrA, respectively. Initial transcriptomic analysis revealed that members of the PE/PPE protein family remained abundantly expressed, even when these three TCRS systems were mutated in Mtb. Further, protein–protein interaction (PPI) analysis was conducted among 16 common genes, ultimately identifying the PPE4 protein as a significant drug target. PPE4, an outer membrane protein from a unique PPE gene family, was chosen as the target protein (PDB ID: 6UUJ) for our machine learning (ML) model to predict novel bactericidal compounds. We trained Convolutional Neural Network (CNN) and Recurrent Neural Network (RNN) models on a curated dataset of drug-like molecules using one-hot encoded SMILES representations, allowing us to predict five promising candidate compounds (Compounds 1–5) with potential activity against TCRS-associated targets. All the predicted compounds demonstrated low IC50 values &lt; 100&#xa0;nM and synthetic Accessibility Scores (SAScore) ≤ 10; notably, only Compound 3 and Compound 4 emerged as potential drug candidates. We further examined the hit compounds using Gaussian to evaluate essential frontier orbital parameters (HOMO–LUMO gap, hardness, and softness). Molecular docking analysis revealed that Ertapenem exhibits the highest binding affinity among the resistance antibiotics, while our model predicted that Compound 3 and Compound 4 arose as promising compounds compared to the remaining resistance antibiotics in terms of binding affinity. Moreover, molecular dynamics simulations were carried out for 100&#xa0;ns. The predicted compounds hold promise for future optimization and experimental testing as potential inhibitors of TCRS-linked survival mechanisms in Mtb.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Transcriptomic dynamics of deletion mutant two-component regulator system of Mycobacterium tuberculosis and machine learning driven novel therapeutic prediction targeting PPE4 protein

  • 期刊论文

  • 发表日期：

    2025年7月9日

查看全文
  • Md. Al Amin,
  • Md. Rashedur Rahman Refat,
  • Sirajul Islam,
  • Shamima Akhtar Sharmin,
  • A. K. M. Mohiuddin,
  • Shahin Mahmud

摘要

Mycobacterium tuberculosis (Mtb) is a rod-shaped, non-motile, and entirely aerobic bacterium, which is responsible for Tuberculosis. Mtb’s remarkable ability to adapt and survive in hostile conditions makes it one of the most resilient and drug-resistant pathogens. One of the key systems that supports this adaptability is the Two-Component Regulatory System (TCRS), which helps Mtb respond to various environmental stresses. In this study, we explored the GSE6750, GSE182749, and GSE53640 microarray datasets of Mtb, specifically focusing on the deletion mutant mprAB, regX3, and MtrA TCRS. Initially, we characterized the primary transcriptome data, resulting in 125 dysregulated genes (DGs) for mprAB, 388 DGs for regX3, and 250 DGs for MtrA, respectively. Initial transcriptomic analysis revealed that members of the PE/PPE protein family remained abundantly expressed, even when these three TCRS systems were mutated in Mtb. Further, protein–protein interaction (PPI) analysis was conducted among 16 common genes, ultimately identifying the PPE4 protein as a significant drug target. PPE4, an outer membrane protein from a unique PPE gene family, was chosen as the target protein (PDB ID: 6UUJ) for our machine learning (ML) model to predict novel bactericidal compounds. We trained Convolutional Neural Network (CNN) and Recurrent Neural Network (RNN) models on a curated dataset of drug-like molecules using one-hot encoded SMILES representations, allowing us to predict five promising candidate compounds (Compounds 1–5) with potential activity against TCRS-associated targets. All the predicted compounds demonstrated low IC50 values < 100 nM and synthetic Accessibility Scores (SAScore) ≤ 10; notably, only Compound 3 and Compound 4 emerged as potential drug candidates. We further examined the hit compounds using Gaussian to evaluate essential frontier orbital parameters (HOMO–LUMO gap, hardness, and softness). Molecular docking analysis revealed that Ertapenem exhibits the highest binding affinity among the resistance antibiotics, while our model predicted that Compound 3 and Compound 4 arose as promising compounds compared to the remaining resistance antibiotics in terms of binding affinity. Moreover, molecular dynamics simulations were carried out for 100 ns. The predicted compounds hold promise for future optimization and experimental testing as potential inhibitors of TCRS-linked survival mechanisms in Mtb.