• Comparative In Silico Assessment of G6PD Variants G163S and R459L Reveals Diagnostic Value of Pathogenicity Prediction Tools
  • Javad Yaghmoorian Khojini,1 Babak Negahdari,2 Mohammad Ali Mazlomi,3,*
    1. 1- Department of Medical Biotechnology, School of Advanced Technologies in Medicine, Tehran University of Medical Sciences, Tehran, Iran 2-Student’s Scientific Research Center, Tehran University of Medical Sciences, Tehran, Iran
    2. 1- Department of Medical Biotechnology, School of Advanced Technologies in Medicine, Tehran University of Medical Sciences, Tehran, Iran
    3. 1- Department of Medical Biotechnology, School of Advanced Technologies in Medicine, Tehran University of Medical Sciences, Tehran, Iran


  • Introduction: Introduction Glucose-6-phosphate dehydrogenase (G6PD) is a critical enzyme involved in the oxidative branch of the pentose phosphate pathway, responsible for the production of NADPH and ribose-5-phosphate. These molecules are essential for redox balance and nucleotide biosynthesis, particularly in erythrocytes. Mutations in the G6PD gene are associated with G6PD deficiency, a common X-linked enzymopathy affecting over 400 million people worldwide. Among various mutations, G163S (also known as G6PD Mahidol) is relatively prevalent in Southeast Asian populations, while R459L is a less characterized variant with reported clinical significance. Although these mutations are documented, the comparative evaluation of their structural and functional impacts using multiple in silico tools remains limited. The present study aims to investigate these two variants using a comprehensive computational pipeline and further explores the agreement and diagnostic value of commonly used pathogenicity prediction tools.
  • Methods: Methods The canonical G6PD protein sequence was retrieved from UniProt (ID: P11413). Two clinically relevant missense variants, G163S and R459L, were selected for analysis. Pathogenicity was assessed using PredictSNP, an integrated meta-server combining SIFT, PolyPhen-1 and 2, SNAP, PhD-SNP, and MAPP. Structural and dynamic consequences were evaluated through DynaMut, which estimates ΔΔG (change in Gibbs free energy) and ΔΔSvib (vibrational entropy) based on normal mode analysis. Residue-level effects were visualized using HOPE. Domain mapping and functional site annotation were conducted using InterProScan, while STRING was used to identify key protein–protein interactions. Reactome was employed to contextualize the functional role of G6PD in cellular pathways.
  • Results: PredictSNP meta-analysis revealed that R459L was consistently predicted to be pathogenic by most tools, including PhD-SNP (86% confidence) and SIFT (71%), aligning with its known clinical relevance. In contrast, G163S displayed a more ambiguous profile: PhD-SNP predicted it as deleterious (68%), while SIFT showed lower confidence (43%). Notably, some tools like SNAP and MAPP predicted it to be neutral, suggesting a moderate functional impact. This discrepancy provided a unique opportunity to evaluate the sensitivity and reliability of individual tools, particularly when clinical data are limited. DynaMut analysis showed that both mutations induced changes in protein stability and flexibility. R459L had a slightly stabilizing global ΔΔG value (+0.194 kcal/mol), but individual models such as mCSM and ENCoM predicted mild destabilization. It also caused a minimal increase in flexibility (ΔΔSvib = +0.010 kcal/mol). In contrast, G163S displayed a positive overall ΔΔG (+0.143 kcal/mol) but with consistent predictions of destabilization across individual tools, along with reduced flexibility (ΔΔSvib = -0.563 kcal/mol), potentially affecting protein folding dynamics. HOPE analysis confirmed that R459L leads to the loss of a positively charged arginine involved in salt bridges and hydrogen bonds, which may disrupt G6PD’s oligomeric structure. G163S, on the other hand, involves a substitution from glycine (a highly flexible residue) to serine, which is bulkier and may hinder helix packing, particularly in the NADP-binding domain. STRING analysis identified strong interactions of G6PD with enzymes like PGD, PGLS, HK1, GPI, and GAPDH, emphasizing the role of G6PD in maintaining redox homeostasis and cellular metabolism. InterPro domain mapping showed that both variants lie within conserved domains critical for enzymatic function. Reactome analysis reaffirmed the involvement of G6PD in the pentose phosphate pathway, highlighting its role in oxidative stress response and NADPH production.
  • Conclusion: This study presents a multi-tool, in silico comparison of two clinically significant G6PD variants, G163S and R459L. While both mutations affect conserved regions, their predicted pathogenicity and structural consequences differ, as reflected by the variability across bioinformatics tools. R459L showed consistent agreement among tools and correlated well with clinical annotations, validating the predictive capacity of tools like PhD-SNP and HOPE. In contrast, the mixed predictions for G163S underscore the necessity of using multi-tool strategies for comprehensive variant interpretation, especially in cases with limited clinical data. The results emphasize that the predictive accuracy of individual tools can vary significantly, and integrating multiple analyses provides a more robust framework for mutation assessment. This comparative approach not only contributes to a deeper understanding of G6PD deficiency mechanisms but also highlights the diagnostic potential of combining computational tools for reliable variant classification.
  • Keywords: G6PD deficiency, missense mutation, in silico analysis, protein stability, pathogenicity prediction