The GUSB sequence was extracted in FASTA format from the UniProt repository (UniProt ID: P08236). A collection of non-synonymous SNPs was generated using information obtained from the PubMed literature search and databases such as dbSNP (Sherry et al., 2001), HGMD (Stenson et al., 2009), ClinVar (Landrum et al., 2014), and Ensembl (Hubbard et al., 2002). The list was cleared to eliminate redundant nsSNPs. The protein data bank provided the crystal structure of the human GUSB gene. Out of the total Data, 1305 Missense variants along with 3’UTR and 5’ UTR were obtained from dbSNP and Ensembl, and are shown in Figure 1. We used a comprehensive computational approach to anticipate harmful mutations in the GUSB at both structural and functional levels, as illustrated in Figure 2.

Sorting Intolerant from Tolerant (SIFT) was used to predict how possible amino acid alterations may affect protein function. SIFT is expanded to include predictions for frame-shifting indels. Human genetic research has extensively used SIFT to assess amino acid substitutions (e.g., cancer, Mendelian disorders, and viral infections). Studies of human diseases and other study subjects are not the only areas in which SIFT is useful. The consequences of missense mutations on model animals such as rats, dogs, and Arabidopsis, as well as agricultural plants, have been studied using SIFT (Sim et al., 2012). If the SIFT score is less than or equal to 0.05, then the mutation is not tolerable (Ng and Henikoff, 2003). The SIFT method was utilized to forecast the impact of these nsSNPs on the protein.

Polymorphism Phenotyping v2 (PolyPhen-2), a program and Web server accessible, estimates the potential effects of amino acid alterations on the stability and activity of human proteins. SNPs are functionally annotated; coding SNPs are mapped to gene transcripts; protein sequence annotations are extracted; and conservation profiles are constructed. The likelihood that the missense mutation would cause harm is then calculated using a combination of all these characteristics (Adzhubei et al., 2013). In addition to estimating the likelihood that a mutation is harmful when it is not, PolyPhen-2 also provides predictions of false-positive and true-positive rates. PolyPhen-2 computes the naive Bayes posterior probability that a particular mutation is harmful. The qualitative assessment of a mutation determines whether it is benign, perhaps harmful, or likely harmful (Adzhubei et al., 2010).

Mutation Assessor (http://mutationassessor.org/r3/) is a sequence-based tool that predicts the functional impact of an nsSNPs on a protein. This server predicts the functional effects of missense polymorphisms and amino acid modifications observed in cancer-causing mutations in proteins. Depending on the affected amino acids and the degree of evolutionary preservation in protein homologs, the functional impact is evaluated (Seker Yilmaz et al., 2021). A sizable collection of polymorphic and disease-associated (OMIM) variants-60,000 in total has been used to verify the technique.

Function Analysis through Hidden Markov Models (FATHMM) is a method that helps predict how both coding and non-coding variations in the human genome affect function. To determine if SNPs and other mutations tend to be damaging, hidden Markov models are employed. When it comes to genetic research and personalized therapy, FATHMM is highly beneficial, as it helps to clarify the significance of genetic variants (Shihab et al., 2013).

A computer method called Mutation Cutoff Scanning Matrix (mCSM) is used to predict how mutations may affect small-molecule binding, protein-protein interactions, protein-nucleic acid interactions, and protein stability. mCSM is an innovative tool for assessing non-synonymous mutations that predicts destabilizing mutations using a graph-based method. If a mutation results in a mCSM score ($\Delta \Delta G$) less than 0, it affects protein structure (Choudhury et al., 2021). It assists in forecasting the impact of mutations on drug binding, hence contributing to the development of more potent medications. It helps understand how changes in protein interactions and stabilization lead to conditions caused by mutations. Directs the creation of proteins for use in industry and medicine that have the appropriate stability and interaction characteristics (Pires et al., 2016).

The free energy change on protein unfolding is estimated using MAESTRO (https://pbwww.che.sbg.ac.at/maestro/web), a multi-agent stability prediction tool. It determines how a point mutation affects the protein's stability by calculating the difference in free energy ($\Delta G$) between the mutant and wild-type proteins. The stability of the protein is changed by a mutation if its score is less than zero (Laimer et al., 2015). Mutations may be limited to certain amino acid classes, exposed or buried residues, and user-specified areas, contingent upon the purpose of the research (Laimer et al., 2015).

PremPS calculates changes in the unfolding Gibbs free energy to assess the effects of single mutations on protein stability. This approach requires a protein's three-dimensional structure (Chen et al., 2020). The PREMPS software determines the mutation's corresponding change in free energy ($\Delta \Delta G$). PREMPS forecasts the impact of the mutation on protein stability based on the calculated $\Delta \Delta G$. Generally, a destabilizing mutation is indicated by a high $\Delta \Delta G$ value, whereas a stabilizing mutation is suggested by a negative $\Delta \Delta G$ (Chen et al., 2020).

DynaMut forecasts the effects of mutations on protein stability by considering MD simulation outcomes. The protein structure is stabilized or destabilized by a mutation based on the calculation of metrics such as changes in free energy ($\Delta \Delta G$). This prediction is essential for evaluating how mutations in biological systems will affect functionality (Rodrigues et al., 2018). PhD-SNP is a tool designed to help understand the genetic basis of illnesses. It analyzes SNPs by assessing their possible negative consequences using a variety of structural and evolutionary factors. A prediction score that indicates the probability of an SNP being deleterious is produced by PhD-SNP. A greater score denotes a higher likelihood that the SNP would result in functional alterations or increased susceptibility to illness, with a score range from 0 to 1 (Capriotti et al., 2006).

A bioinformatics technique called SNP & GO (http://snps-and-go.biocomp.unibo.it/snps-and-go/) integrates data from both Gene Ontology (GO) and SNP annotations to forecast the functional effects of SNPs. This tool facilitates understanding of genomic data in relation to disease and phenotype by illuminating the potential effects of genetic variants on protein function and biological mechanisms. The algorithms used here categorize SNPs as neutral or deleterious (likely to impair protein function) based on the attributes that were retrieved. SNP & GO produces prediction scores determined by the machine learning model that show how likely it is that an SNP will have negative effects. Higher scores indicate a greater likelihood of functional impact, which helps researchers choose SNPs for further study (Capriotti et al., 2013).

MutPred2 was created to predict how missense mutations in human proteins might affect protein function. It evaluates how mutations influence protein structure, function, and interactions by combining diverse features and machine learning methods, delivering significant insights into their potential roles in disease processes (Choudhury et al., 2021).

A technique to determine the aggregation, disorder, helix, and strand propensities resulting from mutations is called Solubility based on Disorder and Aggregation (SODA) (http://protein.bio.unipd.it/soda/). The PDB format structure file or the protein sequence may be entered into this program. SODA uses Fells, ESpritz-NMR, PASTA 2.0, and other resources to forecast mutations of many types, including insertions, deletions, substitutions, and duplications. Based on how differently the WT and mutant proteins are soluble, SODA assigns a final score (Paladin et al., 2017).

Consurf (https://consurf.tau.ac.il/) is a bioinformatics tool designed to find areas in protein sequences that have been conserved across time. It uses evolutionary links among various species and sequence homology to predict the structural and functional significance of amino acid residues. Using multiple sequence alignment, the ConSurf tool was used to assess residue conservation at a particular location (Ashkenazy et al., 2016).

Mutation Assessor categorizes variants according to their predicted functional impact. Moderate-impact mutations are expected to slightly alter protein stability or activity, whereas high-impact mutations are predicted to severely impair protein function, potentially leading to loss of activity or altered biological behavior. Similarly, FATHMM classifies variants into two principal categories based on their anticipated effects: tolerated and damaging. Tolerated variants are predicted to have little or no functional consequence and are likely benign or neutral. In contrast, damaging variants are expected to significantly disrupt protein stability, activity, or molecular interactions, thereby increasing their likelihood of being pathogenic (Shihab et al., 2015). These predictive frameworks enable researchers to prioritize high-risk variants for experimental validation and clinical investigation, particularly those identified as deleterious for disease-association and functional studies.

A total of 449 nsSNPs in the human GUSB gene were analyzed. Sequence-based predictions identified 277, 262, 160, and 192 deleterious mutations using SIFT, PolyPhen-2, FATHMM, and Mutation Assessor, respectively. Structural stability analysis revealed 13, 40, 1, and 11 destabilizing mutations as predicted by mCSM, MAESTROweb, PremPS, and DynaMut, respectively (Figure 3). Overall, this comprehensive multi-tool strategy strengthens confidence in variant prioritization and provides a reliable foundation for subsequent functional validation and disease-correlation studies (Supplementary Table S1).

In addition to sequence-based analyses, four structure-based stability prediction tools DynaMut, PremPS, mCSM, and MAESTROweb were employed in this study. These programs estimate changes in folding free energy ($\Delta \Delta G$) using atomic coordinates derived from the wild-type protein’s PDB structure. Most of these tools integrate biophysical principles with machine-learning or neural-network-based approaches to evaluate the impact of amino acid substitutions on protein stability and conformational dynamics. To enhance prediction reliability and minimize false positives, only those variants consistently identified as deleterious by all four sequence-based tools and all four structure-based methods were selected for further analysis. This stringent integrative screening approach resulted in 15 high-confidence mutations predicted to be both deleterious and structurally destabilizing (Figure 4).

mCSM employs graph-based signatures to predict the impact of amino acid substitutions on protein stability by quantifying changes in folding free energy ($\Delta \Delta G$). Using this approach, mCSM identified 13 destabilizing mutations. MAESTROweb predicted 40 mutations as destabilizing, utilizing a combination of structural descriptors, statistical potentials, and energetic parameters to estimate the effects of mutations on protein stability and functionality. PremPS (Predictions of Protein Stability Changes upon Mutation) identified one destabilizing mutation. This tool integrates structural environmental features with probabilistic energy functions to assess the effects of single-point mutations on protein stability. DynaMut detected 11 destabilizing mutations by combining graph-based signatures with normal-mode analysis to assess their effects on both protein stability and conformational dynamics (Supplementary Table S2).

To further investigate the disease associations of the prioritized nsSNPs in the human GUSB gene, we employed three well-established pathogenicity prediction tools: PhD-SNP, SNPs&GO, and MutPred. These computational platforms classify variants as benign or pathogenic based on probability scores derived from sequence features, evolutionary conservation, and functional annotations, thereby enabling reliable assessment of disease association. From our integrative sequence- and structure-based analyses, 15 nsSNPs were identified as high-confidence deleterious and destabilizing variants. These shortlisted mutations were subsequently subjected to pathogenicity evaluation using the three prediction tools.

PhD-SNP classified 12 out of the 15 variants (80%) as pathogenic, indicating a strong likelihood of disease association. Similarly, SNPs&GO predicted 10 of the 15 variants (66.66%) to be disease-related, further supporting the potential clinical relevance of these mutations (Table 1). To enhance confidence and minimize false-positive predictions, we compared results from PhD-SNP and SNPs&GO. Eight variants were consistently identified as pathogenic by both tools, namely: L214P, L566P, N135D, P108L, R26L, T180P, Y388C, and Y399C. These eight mutations were therefore prioritized as the most probable disease-causing variants and selected for subsequent detailed analysis.

Analysis of amino acid conservation within a protein structure provides critical insight into the functional and structural importance of specific residues and reflects the evolutionary constraints that shape them. Highly conserved residues are often essential for maintaining structural integrity, catalytic activity, or molecular interactions. In general, the likelihood of mutation tolerance is inversely related to the degree of conservation; residues that are strongly conserved across species are less likely to accommodate substitutions without functional consequences.

To evaluate evolutionary conservation in the human GUSB protein, we performed a ConSurf analysis (Figure 5). The results indicated that residues within the regions 360-370, 380-390, 415-425, and 566-571 exhibit a high degree of conservation compared to other segments of the protein. These conserved regions likely play critical roles in maintaining structural stability and functional competence. Furthermore, ConSurf classification revealed that several of these highly conserved residues are buried within the protein core, suggesting a structural role in maintaining stability, while others are exposed on the protein surface, potentially contributing to functional interactions. These findings reinforce the biological significance of mutations occurring within conserved regions of the GUSB protein.

Amino acid solubility plays a critical role in maintaining proper protein folding and function (Balch et al., 2008). Previous studies have demonstrated that reduced protein solubility can promote misfolding and aggregation, which are central mechanisms underlying several neurodegenerative and aggregation-related disorders (Knowles et al., 2014). Insoluble protein regions tend to cluster, leading to pathological aggregates associated with diseases such as Parkinson’s disease, Alzheimer’s disease, and amyloidosis (Thal et al., 2015).

SNPs are the most common form of genetic variation and are widely associated with diverse human diseases. Comprehensive analysis of these variants provides valuable insight into the molecular mechanisms underlying disease progression and may facilitate the development of targeted therapeutic strategies. In this study, we focused on nsSNPs in the GUSB gene, a critical lysosomal hydrolase responsible for glycosaminoglycan degradation.

To evaluate the effects of prioritized variants on the solubility of the GUSB protein, we used the SODA tool. SODA provides a comprehensive assessment by estimating mutation-induced changes in aggregation propensity, intrinsic disorder, helix propensity, and strand propensity. This integrative analysis enables a deeper understanding of how specific amino acid substitutions may influence protein folding behavior and aggregation potential (Choudhury et al., 2021).

From our earlier pathogenicity assessment, eight nsSNPs were identified as disease-associated variants (Table 2). These eight mutations were subsequently analyzed using SODA to determine their impact on protein solubility. Among them, three variants were predicted to reduce protein solubility. Decreased solubility is typically associated with increased aggregation propensity, which may contribute to disease pathogenesis by forming aberrant protein assemblies. These findings suggest that these three variants could potentially promote aggregation-mediated dysfunction of the GUSB protein.

In contrast, the remaining five nsSNPs were predicted to enhance protein solubility. Increased solubility generally reflects a reduced tendency to aggregate and may help maintain a more stable, functionally competent protein conformation. Such mutations may therefore have a comparatively lower risk of inducing aggregation-related pathogenic effects. The analysis demonstrated that three of the eight pathogenic mutations (37.5%) are predicted to decrease protein solubility and increase aggregation tendency. This observation is particularly significant, as protein aggregation is a well-established contributor to several pathological conditions, including neurodegenerative and aggregation-related disorders. In contrast, the remaining variants were predicted to maintain or enhance solubility, suggesting comparatively lower aggregation risk.

Overall, the SODA analysis expands our understanding of how specific GUSB variants may influence protein solubility and aggregation behavior. This comprehensive computational analysis highlights a subset of GUSB nsSNPs with strong evidence of deleterious, destabilizing, and pathogenic effects. By integrating assessments of stability, pathogenicity, and solubility, the study provides a detailed molecular perspective on how specific variants may impair $\beta$-glucuronidase function and contribute to disease pathogenesis.

To gain deeper insight into the molecular consequences of the identified pathogenic variants, we performed a detailed structural and conformational analysis of the GUSB protein, with particular emphasis on residues 566 and 388. A comparative evaluation of the wild-type and mutant structures revealed notable alterations in intermolecular interactions, highlighting the structural impact of these substitutions.

At position 566, the wild-type (Leu) protein exhibited two conventional hydrogen bonds, whereas the mutant (Pro) form retained only one, indicating a reduction in stabilizing interactions (Figure 6A). Two weak hydrogen bonds were preserved in both forms, and neither structure displayed water-mediated hydrogen bonds. Additionally, halogen bonds, ionic interactions, and metal-coordinated contacts were absent in both the wild-type and mutant configurations.

More pronounced differences were observed in non-covalent stabilizing interactions. Aromatic interactions, which play a critical role in maintaining structural stability, were present in the wild-type structure (eight interactions) but were completely absent in the mutant. Similarly, hydrophobic interactions, essential for maintaining protein core integrity, were dramatically reduced from 17 in the wild type to only 1 in the mutant. No carbonyl interactions were detected in either form.