A comprehensive in silico investigation into the deleterious nonsynonymous single nucleotide polymorphisms of the human transcription factor EB (TFEB) gene and their predicted association with cancer

Highlights

• •Integrated bioinformatics approaches analyzed TFEB nsSNP variants.
• •A large-scale screening evaluated 20,024 nsSNPs in the TFEB gene.
• •Six variants were predicted to be deleterious and structurally destabilizing.
• •Mutations altered TFEB protein stability and its DNA-binding interaction.
• •Identified variants may guide future functional and clinical studies.

Abstract

Transcription factor EB (TFEB) is an essential protein that is connected to a number of diseases, such as lysosomal storage disorders and cancer. Patients with glioblastoma multiforme and uterine endometrioid carcinoma have already been identified to have nonsynonymous mutations in this gene. These mutations may have an impact on TFEB protein’s functions and structure. Therefore, to find possible biomarkers for different disease treatments, the most harmful single nucleotide polymorphisms (SNPs) of the TFEB protein have been identified in this study. The goal is to create a systematic dataset of the SNPs related to the TFEB gene, which could be useful in the diagnosis and management of many disorders linked to the target gene. The SNPs of the TFEB protein were analyzed via a wide range of bioinformatics techniques, including both sequence and structure-based methodologies. Nonsynonymous SNV research can be advanced through the application of various machine learning methods that have been developed as a result of recent advancements in computational platforms. Among 449 nsSNPs, a total of 6 nsSNPs have been found to be harmful, destabilizing, and disease-causing. Each nsSNP interferes with its function. All of the mutant proteins interact with the DNA molecule during docking more effectively than the alphafold (wild type) protein does. A strong correlation between cancer and mutations in R315 has been identified. The detrimental effects of nsSNPs and noncoding SNPs on the structure and activities of proteins will help researchers understand the crucial role that mutations play in the molecular pathways involved in a variety of disorders. The discovery of possible targets for the diagnosis of diseases and treatment interventions will eventually result from this. Moreover, this thorough analysis can facilitate the exploration of potential disease-causing SNPs in the TFEB gene and assist in identifying effective drugs or pharmacological targets. Consequently, further experimental mutational research, genome-wide association studies, and clinical-based studies are essential to validate these findings.

1. Introduction

TFEB, along with MITF, TFE3, and TFEC, are the four uniquely encoded genes that make up the microphthalmia (MiT) family of transcription factors.¹ TFEB is classified as a basic helix-loop-helix-leucine-zipper transcription factor (bHLH-ZiP).² The basic portion of MiT proteins is the same and is required for DNA binding. The HLH and Zip sections are essential for dimerization and are quite similar. However, they differ significantly outside of these regions.² The transactivation domains and basic DNA-contacting domains of all family members recognize similar DNA sequences, suggesting possible commonality in their target gene repertoire. These components can also heterodimerize with one another. For transcriptional activation, TFEB also has a conserved activation domain.3, 4 MiT transcription factors can bind DNA as both homodimers and heterodimers when bound to another family member.⁵

Initially, lysosomes were defined as stationary organelles that were responsible for the final breakdown of cell garbage. Recent findings suggest that precisely regulated transcriptional control affects both lysosomal biogenesis and function, casting doubt on this theory. Microarray analysis revealed that genes encoding lysosomal proteins are co-expressed in many cell types and environments.⁶

A palindromic region (TCACGTGA) that mimics an E-box is shared by lysosomal genes, as shown by promoter analysis; this region is referred to as the coordinated lysosomal expression and regulation (CLEAR) motif.⁶ TFEB was shown to directly bind to CLEAR elements, upregulating the expression of the entire gene network known as the CLEAR network, which is composed of genes with the CLEAR regulatory motif in their promoter.⁷ Increased lysosomal activity is caused by TFEB overexpression, which also increases lysosomal enzyme levels and lysosomal numbers.⁶ Further studies have shown that TFEB modulates the expression of a broader group of genes, including genes involved in lysosomal development and function, autophagy, and lysosomal exocytosis.⁷ TFEB binds to the promoter regions of multiple autophagy genes, triggering autophagosome biogenesis and autophagosome‒lysosome fusion events. Lipid droplets and damaged mitochondria are cleared as a result of TFEB overexpression. To increase intracellular clearance and regulate primary cellular degradative pathways, TFEB coordinates a transcriptional program.8, 9, 10 Instead of controlling its targets' baseline transcription, TFEB increases their transcriptional levels so that they can react to external stimuli.

Both fetuses and adults express large amounts of the transcription factor EB, which has limited tissue and cellular specificity.¹¹ Under nutrient-rich, resting cell circumstances, TFEB is mostly cytosolic and inactive. When lysosomal malfunction or hunger occurs, TFEB quickly enters the nucleus and initiates gene transcription.6, 8 The phosphorylation status of TFEB essentially determines its cellular localization and function. Extracellular signal regulated kinase 2 (ERK2, also known as MAPK1) and mammalian target of rapamycin complex 1 (mTORC1) are the main protein kinases that cause the phosphorylation of TFEB in nutritious conditions in the majority of cell types. Both of these enzymes are master regulators of cellular growth.12, 13 In response to receptor activator of nuclear factor κB ligand (RANKL), protein kinase Cβ (PKCβ) phosphorylates TFEB in its C-terminal region in osteoclasts.¹⁴ Ras-related GTP-binding or Small Rag GTPases deliver mTORC1 to the lysosomal membrane, where it enhances its activation via the small GTPase Rheb.¹⁵ Additionally, TFEB is bound by active Rag GTPases, which bring it to the lysosomal membrane and encourage mTORC1 to phosphorylate it.¹⁶ In obese mice, TFEB depletion in the liver impairs hepatic catabolism and exacerbates metabolic imbalance; in contrast, TFEB overexpression reverses this effect, rescuing obesity and related metabolic syndrome.¹⁰

A key modulator of osteoclast activity, receptor activator of nuclear factor κB ligand (RANKL), controls TFEB function in osteoclasts. Osteoclast-specific TFEB deletion results in increased bone mass and decreased osteoclast function, confirming that TFEB regulates bone resorption in this tissue.¹⁴ A compromised T-cell-dependent antibody response results from the simultaneous reduction of TFEB and TFE3 (members of the MiT family).¹⁷ Major histocompatibility complexes (MHCs) in dendritic cells present antigens. TFEB has been shown to regulate antigen presentation, demonstrating a broad role for TFEB in immune response activation.¹⁸ Lysosomal storage disorders (LSDs) are a class of disease in which TFEB overexpression has been demonstrated to be beneficial. These disorders occur when certain lysosomal proteins present genetic abnormalities that cause substrates to accumulate within the lumen of lysosomes.¹⁹ Amplification of TFEB has been shown to be beneficial. It reduces substrate accumulation in cellular and murine models of multiple types of sulfatase deficiency, mucopolysaccharidosis type IIIA, Batten disease, Pompe disease, Gaucher disease, Tay‒Sachs disease and cystinosis. Additionally, this overexpression lessened the severity of cellular and tissue abnormalities and enhanced overall autophagy and lysosomal function.20, 21, 22

The most common genetic differences detected in the human genome are known as SNPs. They occur approximately every 100–300 base pairs and refer to changes in a single nucleotide at a specific genomic region among individuals.²³ Nonsynonymous SNPs (nsSNPs) are SNPs that alter the amino acid sequences of proteins to produce an effect. Missense mutations induced by these specific SNPs account for approximately half of all genetic disorders.²⁴ Numerous reports have linked this type of genetic variation to specific phenotypes or diseases, such as a predisposition to cancer. Pancreatic adenoductal carcinoma has also been shown to overexpress MITF, TFEB, and TFE3, suggesting that they promote tumor growth by inducing autophagy.²⁵ Interestingly, tumors with overexpressed or amplified MiT-TFE genes exhibit upregulation of RagD, a direct transcriptional target of TFEB, leading to hyperactivation of mTORC1.²⁶ One study revealed a link between functional SNPs in the TFEB gene and an increased risk of several neurological diseases, notably Huntington's, Parkinson's, and Alzheimer's diseases.²⁷ According to other studies, TFEB SNPs are involved in the emergence of uncommon malignant tumors. A unique subtype of juvenile renal cell carcinoma is caused by aberrant expression of TFEB via a particular gene translocation known as alpha-TFEB fusion.²⁸ The development of coronary artery disease (CAD) and acute myocardial infarction (AMI) may be related to altered levels of TFEB-mediated expression of genes and subsequent alterations to the autophagic–lysosomal network.²⁹

Nonsynonymous missense SNPs (nsSNPs) have long been considered harmful. They modify amino acids inside a gene, which can alter the function of the protein. It can be costly and time-consuming to empirically determine the impact of several nsSNPs. However, studying a large number of SNPs existing inside a specific gene is both practical and cost-effective when performing in silico analysis through readily available web-based bioinformatics tools. Significant advancements in the field of disease-related nsSNP analysis have been made via computational genomics.³⁰

While numerous independent programs are accessible, they require programming language proficiency. They rely on machines as well. To guarantee repeatability and reproducibility, web servers were chosen over standalone computational tools for this investigation. To identify the nsSNPs that are most likely to be detrimental, all of the nsSNPs found in the TFEB gene were gathered, filtered and examined.

Our in-silico analysis of deleterious nsSNPs in the TFEB gene seeks to identify variants that may impair its critical regulatory functions in cellular processes such as autophagy and lysosomal biogenesis, which are often dysregulated in cancer. By characterizing these nsSNPs, our study aims to uncover their potential roles in disrupting TFEB-mediated pathways, thereby contributing to the development of precision cancer therapies tailored to specific genetic profiles. Furthermore, this work has the potential to advance personalized medicine by identifying patient-specific TFEB variants that could influence treatment responses, enabling more effective therapeutic strategies. Additionally, our analysis may reveal novel therapeutic targets, paving the way for innovative pharmacological interventions in cancer and related diseases.

2. Materials and methods

2.1. Retrieval of data

All the SNPs in the TFEB gene, as well as the necessary information (such as location, altered amino acid residue, reference SNP ID), were gathered from the dbSNP database of the National Center for Biotechnology Information (NCBI) (https://www.ncbi.nlm.nih.gov/snp/)³¹. Although additional web-based tools such as Ensembl, gnomAD, ClinVar, VarSome, SNPnexus, and MutationTaster were also explored, all identified variants were fully represented within the dbSNP database. Therefore, to maintain conciseness and avoid redundancy, dbSNP was reported as the primary source of SNP data in this study. From UniProt databases (https://www.uniprot.org/)³² we have gathered protein accession number, amino acid sequences. Only the nsSNPs were taken into account for further investigation. The overall workflow of the study has been shown in Fig. 1.

Fig. 1.

Overall workflow for identifying deleterious nsSNPs of the TFEB gene.

2.2. Identification of deleterious SNPs

To identify the most damaging SNPs, the nsSNPs of the TFEB gene were initially screened via the SIFT-Sort Intolerant From Tolerance program (https://sift.bii.a-star.edu.sg/). SIFT analyzes the physical characteristics of individual amino acids as well as sequence homology with other comparable proteins to determine whether amino acid replacement impacts protein functions. The basis of SIFT is the concept that there is an association between protein function and evolution. Prediction scores that were greater than 0.05 were regarded as tolerant, whereas those that were less than or equal to 0.05 were deemed affected.³³

2.3. Analyzing the effects of deleterious SNPs

Two additional tools were used to anticipate the functional consequences of the SIFT predictions. PolyPhen-2 (phenotyping polymorphism version 2) is a technology that uses machine learning to forecast how nsSNPs affect a protein's structure and function.³⁴ Polyphen-2 anticipates the potential impact of each substitution on a protein's structural and functional characteristics. The structural and evolutionary conservation analysis yielded this prediction. Polyphen-2, in contrast to SIFT, typically provides three prediction categories: “Most likely, Damaging,” “Possibly Damaging,” and “Benign.” Polyphen-2 provides a score that ranges from 0 to 1.³⁵ PolyPhen-2 employs multiple sequence alignment, first utilizing a clustering technique to identify the homologous sequences for study and then building and honing their multiple alignments. The naive Bayes classification system uses each attribute to predict the functional importance of an amino acid substitution.³⁶

The functional impact of protein sequence variations, such as minor insertions and deletions as well as single amino acid changes, is predicted by Provean (Protein Variation Effect Analyzer) (http://provean.jcvi.org/index.php).³⁷ The prediction is based on the shift in a query sequence's similarity to a group of related protein sequences caused by a specific variation. The method needs to determine a pairwise semiglobal sequence alignment score between the targeted sequence and each of the associated sequences to make this prediction.³⁷ If the PROVEAN score is equal to or less than a predetermined threshold (−2.5), the protein variant is likely to have a “deleterious” effect; a score greater than the threshold is predicted to have a “neutral” effect. A lower score threshold (such as −4.1) can also be utilized to improve the specificity of detection or the confidence with which harmful variations are found.³⁷

2.4. Identification of disease-related SNPs in the TFEB gene

To determine whether SNPs are linked to disease, the web tools SNPs&GO and PhD-SNP were used. Using the support vector machine method, the connection between SNPs and diseases is predicted by SNPs and GO at 81% accuracy.38, 39 A likelihood value greater than 0.5 indicates a relationship between SNPs and medical conditions.⁴⁰ Additionally, PhD-SNPs evaluate the relationships between SNPs and diseases, categorizing them as neutral or disease-associated on the basis of a reliability index score ranging from 0 to −9. The risk of a disease is connected with a value >0.5, whereas a value ≤0.5 is regarded as neutral.³⁹ With the aid of both of these tools, the deleterious nsSNPs that were predicted in the preceding section were further investigated for correlations with human diseases. On the basis of how likely they are to result in human diseases, the SNPs were divided into two categories: “Disease” and “Neutral.” For further assessment, the SNPs classified as “Disease” were sorted.

2.5. Protein stability prediction

I-Mutant 2.0 and MUpro, two separate tools, were utilized to examine whether changes in amino acid composition had an impact on the stability of proteins. I-Mutant 2.0 is an algorithm based on support vector machines (SVMs) that automatically anticipate changes in protein stability caused by modified residues (such as SNPs). The unfolding Gibbs free energy value of the wild-type protein was deduced from the unfolding Gibbs free energy value of the mutant protein (unit kcal/mol), and a numerical approximation was provided of the change in free energy (DDG or delta delta G value). DDG > 0 denotes increased protein stability, whereas DDG < 0 denotes a decrease in protein stability.⁴¹ Depending on whether structural or sequencing information is used, the I-Mutant accurately predicts 80% or 77% of the dataset.⁴² All submissions were performed at pH 7.0 and 25°C.

Protein stability is also frequently evaluated via the MUpro server.43, 44 It uses a neural network in addition to an SVM for calculations. These methods calculate the impact of a single amino acid modification on the stability of a protein via a positive or negative score and report the results as increasing or decreasing. A score smaller than zero indicates a deterioration in protein stability.⁴⁵

2.6. Prediction of protein structural alterations and loss of activity

The tool MutPred2 (http://mutpred.mutdb.org/) was utilized to predict modifications in several protein characteristics, such as changes in binding and activity. It was trained by using databases of harmful and unlabeled variants from the Human Gene Mutation Database (HGMD), dbSNP, SwissVar and interspecies pairwise alignment. This website can predict more than 50 protein properties. MutPred2 uses a neural network technique to predict how amino acid changes may affect phenotypes.⁴⁶ The effects could include changes in the stability and structure of the protein, disruptions to macromolecular binding, the removal of a PTM site, etc., all of which could result in significantly altered protein phenotypic characteristics. The application offers various likelihood percentages for distinct qualities, hence highlighting noteworthy SNPs of TFEB. This tool can be utilized to examine multiple facets of protein activity in human tissue that can be influenced by SNPs.

2.7. Predicting phenotypic effects of the nsSNPs

The HOPE (Have (y) Our Protein Explained) (https://www3.cmbi.umcn.nl/hope/) was used to examine the structural and functional implications of the point mutations.⁴⁷ This automated mutant analyzer, a next-generation web tool, generates a report for every mutation that shows how the mutation affects the size, charge, bonding pattern and interaction of the protein with other molecules. These data are processed via an assessment system, which also forecasts how mutation affects protein function and three-dimensional structure. After that, a report is created that details and illustrates how the mutations have affected the protein.⁴⁷ This site also contains 100% conserved residues, which helps interpret mutations that are more likely to cause damage and consequently have a substantial influence on protein function. The inputs for this were the sequence of the protein, the locations of the changes in amino acids, and the modified residues. To learn more about the alterations in protein levels caused by the SNPs, we examined the TFEB SNPs on this server.

2.8. Mutation clusters prediction

Mutation 3D (http://www.mutation3d.org/about.shtml) was used to further evaluate the functionally significant SNPs in the TFEB genes that were predicted via all of the aforementioned techniques and to determine whether the amino acid modifications fit into any cluster that could cause any functional hotspots of mutations. Mutation3D provides information on cluster mutations and illustrates the spatial organization of altered amino acid residues in protein structures.46, 48, 49 This program is able to detect amino acid clusters on protein models and structures via a complete-linkage clustering approach.

2.9. TFEB-protein interaction prediction

To comprehend a protein's function, structure, molecular action, and regulation, one must be aware of its interaction partners. The interaction profile of TFEB with other proteins was determined via STRING (https://string-db.org/). An interaction network with high confidence (score ≥90%) was chosen for the investigation. To better understand the network interactions of proteins, the server links data from other databases and uses computational methods to identify functional and physical correlations.50, 51, 52 The highest confidence score was used when searching for the protein by name.

2.10. Identification of nsSNPs in protein domains

The TFEB protein FASTA sequence has been submitted to the InterPro server (https://www.ebi.ac.uk/interpro/).⁵³ Using several signature qualities, the InterPro server predicts families, conserved domains, and crucial sites of proteins. The domains were identified via the InterPro server, after which the locations of the nsSNPs inside each domain were manually determined.

2.11. Prediction of posttranslational modifications (PTMs)

Posttranslational modifications (PTMs) of proteins govern a variety of biological processes, including signaling cascades and protein‒protein interactions.⁵⁴ Understanding the structure surrounding PTM sites in great detail is important because mutations in residues can cause ambient or orthosteric effects that result in changes in energy conformations and stability. These findings help to elucidate how PTMs affect protein folding.⁵⁵ Phosphorylation, methylation, ubiquitination, acetylation, N-linked glycosylation, and palmitoylation are a few prominent PTMs that are crucial to the study of clinical conditions. We utilized PhosphoSitePlus, a tool designed to investigate posttranslational modifications observed in experimental studies.⁵⁶ Furthermore, we also used iPTMnet, which serves as a portal for interactive and methodical investigations of PTM networks and conservation.⁵⁷

2.12. 3D modeling of proteins

We acquired the TFEB protein FASTA sequence from the UniProt website (https://www.uniprot.org/). SWISS-MODEL, which is a tool for protein homology modeling (https://swissmodel.expasy.org/), was used to model all filtered detrimental nsSNP-containing proteins,58, 59 The structure predicted by Alphafold (https://alphafold.ebi.ac.uk/) was utilized as the template because the full crystal structure of TFEB is not available. This is because Alphafold's anticipated structure is quite accurate. Structure ID: AF-P19484-F1.⁶⁰ An alignment between the predicted Alphafold structure and the X-ray structure was required to validate the use of this structure as a template. To acquire the TM score and RMSD score, TM-Align was utilized.⁶¹ The RCSB PDB (https://www.rcsb.org/structure/7Y62) provides the X-ray structure, which has a modeled residue count of 75 amino acids. The SAVES v6.0 server (https://saves.mbi.ucla.edu/) and the SWISS-MODEL structure assessment system were used to evaluate the quality of the modeled 3D structures.⁶² SAVES v6.0 has five distinct validation methods. Among these, we used ERRAT⁶² and PROCHECK⁶³ to verify the mutant structures we modeled. Following the prediction and verification of each mutant model, they were all examined in TM-align to provide the TM score and the root mean square deviation (RMSD) score.

2.13. Model refinement and evaluation

The TFEB wild-type Alphafold structure, as well as the mutant protein structures, were submitted to GalaxyRefine (https://galaxy.seoklab.org/) for overall structural relaxation via dynamic simulation. Molecular dynamics simulation is used in this method to achieve increased structural relaxation.64, 65, 66 These tools have been used for evaluation: the ERRAT (https://servicesn.mbi.ucla.edu/ERRAT/) and the SWISS-MODELModel Structure Assessment (https://swissmodel.expasy.org/assess). MolProbity, favorite region percentage in the Ramachandran plot, QMEAN, Z scores and ERRAT scores—all of which are required to evaluate the modeled structures.

2.14. Prediction of cancer-associated SNPs

CScape is a combinatorial website that assesses point mutations in both coding and noncoding regions of the genome to determine their carcinogenic potential.67, 68. On the basis of P values between 0 and 1, CScape classifies a mutation as either a cancer driver or a neutral mutation. A value > 0.5 is considered to be oncogenic, whereas a value < 0.5 is considered neutral. We further utilized Dr. Cancer,⁶⁹ FATHMM,⁷⁰ Cancer Genome Interpreter⁷¹ bioinformatics tools to assess the cancer-related association of nsSNPs.

A web-based tool for examining, displaying, and evaluating multimodal cancer genomics data is the cBioPortal for Cancer Genomics (http://cbioportal.org). The gateway distills molecular profile information from cancerous tissues and cell lines into easily interpreted events related to gene expression, proteomics, genetics, and epigenetics.⁷² The CBioPortal server was also used to search for any clinical evidence of the mutations.

2.15. Pathway analysis

The Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway database was utilized to investigate the signaling pathways in which TFEB is functionally involved. KEGG is a widely used bioinformatics resource that integrates genomic, chemical, and systemic functional information, enabling the mapping of genes or proteins onto molecular interaction networks and signaling pathways.73, 74 Using KEGG pathway analysis tools, we identified TFEB-associated pathways to better understand its regulatory role in cellular processes.

2.16. Docking analysis

HADDOCK v2.4 was used to dock the native and mutant TFEB with the E-box⁷⁵ of the DNA molecule. The DNA (E-box) molecules and modeled natural and mutant proteins were utilized. From the literature, we investigated the interaction (binding) residues between TFEB and DNA (E-box).6, 7, 76 Using the web-based DNA sequence-to-structure conversion tool, the most common target site of TFEB, 5′-GTCACGTGAC-3′, was given a 3D structure (in the Protein Data Bank, or PDB) in B-form.⁷⁷

The binding residues in TFEB that were gathered along with those in the companion E-box were utilized as active residues, whereas the residues that were nearby were utilized as passive residues. We employed HADDOCK default parameters in our investigation. Electrostatic energy, van der Waals energy, binding energy, the intervector projection angle restrains energy, symmetry restrains energy, distance restrains energy, diffusion anisotropy energy, the dihedral angle restrains energy, desolvation energy, and buried surface area are among the energy terms that make up the weighted sum (HADDOCK score) used in HADDOCK scoring.⁷⁵ We also computed the electrostatic energy, van der Waals energy, desolvation energy, constraint violation energy and buried surface area in addition to the HADDOCK score. The DNAproDB service was used to visualize the interacting residues between the wild-type (alphafold) and mutant DNA and to analyze intermolecular interactions.⁷⁸

3. Results

3.1. Retrieval of the nsSNP dataset

The NCBI dbSNP database has 20,024 SNPs (intronic, synonymous, nonsynonymous, 3′ and 5′ untranslated region) related to the TFEB gene. Among the 20,024 SNPs, 18,723 (93.5%) were determined to be present in the intron area, 449 were missense or non-synonymous SNPs (2.24%), 246 were synonymous SNPs (1.22%), 480 in the 5′ untranslated region (2.4%), and 275 in the 3′ untranslated region (1.37%). Since nonsynonymous SNPs frequently change the encoded amino acid, only these SNPs were taken into consideration for additional analysis in this study. The percentages of the SNPs in the TFEB gene is shown in Fig. 2.

Fig. 2.

Overview of the mutations found in the TFEB gene from the dbSNP database. Different colors represent different types of mutations.

3.2. Identification of deleterious nsSNPs

A total of 182 of the 449 nsSNPs for which SIFT was predicted to have detrimental impacts on protein function were further examined via downstream methods after receiving a score of less than or equal to 0.05 on a scale of 0--1. The details are shown in Supplementary Table 1.

3.3. Analysis of the effects of deleterious SNPs

Out of 182 nsSNPs, 38 nsSNPs were predicted to be benign, 41 nsSNPs were possibly damaging, and 103 nsSNPs were predicted to be damaging. 99 of these 182 nsSNPs were determined to be deleterious by PROVEAN.The results are shown in Supplementary Table 2.

To ensure that only the most extremely harmful SNPs would be examined, we selected those nsSNPs that scored 0 in SIFT, 1 in PolyPhen-2, and less than −4.1 in PROVEAN. There are eighteen such nsSNPs, which are provided in Table 1.

Table 1.

Prediction of the most deleterious nsSNPs by SIFT, Polyphen-2 and PROVEAN.

Variant ID	Mutation	SIFT		Polyphen-2		PROVEAN
		Score	Result	Score	Result	Score	Result
rs369809425	R464C	0	Affected	1	Probably damaging	−5.524	Deleterious
rs1040783093	P456L	0	Affected	1	Probably damaging	−7.351	Deleterious
rs1303746066	G327V	0	Affected	1	Probably damaging	−8.048	Deleterious
rs768261155	G327S	0	Affected	1	Probably damaging	−5.418	Deleterious
rs371625604	R315H	0	Affected	1	Probably damaging	−4.744	Deleterious
rs748808677	R315C	0	Affected	1	Probably damaging	−7.59	Deleterious
rs144086780	R304C	0	Affected	1	Probably damaging	−4.652	Deleterious
rs779208482	R303C	0	Affected	1	Probably damaging	−6.092	Deleterious
rs1463113136	R271C	0	Affected	1	Probably damaging	−7.606	Deleterious
rs1281647905	D269H	0	Affected	1	Probably damaging	−4.999	Deleterious
rs1561848484	R254C	0	Affected	1	Probably damaging	−7.619	Deleterious
rs1327480600	R234W	0	Affected	1	Probably damaging	−7.973	Deleterious
rs1337401046	A229D	0	Affected	1	Probably damaging	−5.98	Deleterious
rs909431482	Y194C	0	Affected	1	Probably damaging	−8.84	Deleterious
rs374238733	R93W	0	Affected	1	Probably damaging	−4.612	Deleterious
rs1169506393	Y82C	0	Affected	1	Probably damaging	−7.617	Deleterious
rs1212743510	R22W	0	Affected	1	Probably damaging	−5.098	Deleterious
rs1768890982	R465W	0	Affected	1	Probably damaging	−4.347	Deleterious

3.4. Prediction of disease-related SNPs

The PhD-SNP and SNPs&GO servers identified 16 and 9 of the 18 nsSNPs, respectively, as potentially disease-related nsSNPs. The results are shown in Supplementary Table 3. Among these nsSNPs, nine that were predicted to be associated with disease by SNPs&GO were also predicted to be associated with disease by PhD-SNPs. Therefore, we chose these 9 nsSNPs for additional research. These are shown in Table 2.

Table 2.

Prediction of disease-associated nsSNPs by PhD-SNP and SNPs&GO.

Variant ID	Mutation	PhD-SNP		SNPs&GO
		RI	Result	Predictions	RI	Probability
rs369809425	R464C	0	disease	Disease	0	0.114
rs1303746066	G327V	8	disease	Disease	4	0.7
rs768261155	G327S	8	disease	Disease	3	0.627
rs371625604	R315H	9	disease	Disease	0	0.521
rs748808677	R315C	8	disease	Disease	2	0.614
rs1463113136	R271C	9	disease	Disease	5	0.759
rs1281647905	D269H	9	disease	Disease	1	0.572
rs1561848484	R254C	9	disease	Disease	6	0.789
rs1327480600	R234W	9	disease	Disease	2	0.605

3.5. Prediction of protein stability alterations

Six of the nine nsSNPs from the TFEB gene that were predicted by the five tools previously stated to be harmful and associated with disease resulted in decreased protein structural stability, but the remaining three nsSNPs showed increased stability, as predicted by I-Mutant 2.0. Every nsSNP was predicted by MUpro to result in decreased protein structural stability. These data are shown in Supplementary Table 4. Together with the DDG value, reliability index scores were given to help elucidate the decline in protein stability. Sorted SNPs compromise the overall stability of the protein structure in addition to causing phenotypic harm. The nsSNPs that decrease protein stability were further analyzed and are shown in Table 3.

Table 3.

Stability analysis of the nsSNPs of the TFEB protein.

Variant ID	Mutation	I-Mutant 2.0	MUpro
		DDG	DEL(G)
rs369809425	R464C	−0.34	−1.0129419
rs371625604	R315H	−0.9	−1.3631867
rs1463113136	R271C	−0.69	−0.9478444
rs1281647905	D269H	−0.73	−1.27613
rs1561848484	R254C	−0.47	−1.3458621
rs1327480600	R234W	−0.62	−0.6004257

3.6. Prediction of protein structural alterations and loss of activity

The MutPred2 server was utilized for predicting the probability scores and general scores for each of the six nsSNPs of the TFEB protein that were predicted by both I-Mutant 2.0 and MUpro to have lower stability. The MutPred2 server also predicted changes in the structural and functional features of each of the six nsSNPs. Changes in the disordered interface, loss of the B-factor, gain of acetylation, modified DNA binding, modified coiled coil, gain of relative solvent accessibility, loss of ubiquitylation, loss of phosphorylation and modified metal binding are among them. P values and their likelihoods are also given. Among the six nsSNPs, two (R271C and D269H) had scores greater than 0.8, indicating that they have high pathogenic qualities. R254C, another nsSNP, with a score of 0.798, indicating that it cannot be ruled out as nonpathogenic. Given their scores of less than 0.8, the remaining 3 SNPs can be deemed nonpathogenic. These predictions suggest that a number of nsSNPs may cause structural or functional changes in the TFEB protein. All the results are shown in Table 4.

Table 4.

Prediction of alterations in the structure and activity of TFEB by Mutpred2.

Mutation	Mudpred2 score	Alteration	Probability	P Value
R464C	0.536	Loss of Phosphorylation at S467	0.52	3.60E-03
		Loss of Methylation at K460	0.19	8.10E-03
		Loss of Ubiquitylation at K460	0.15	5.00E-02
R315H	0.628	Altered Disordered interface	0.29	0.03
R271C	0.854	Altered Disordered interface	0.34	0.01
		Altered DNA binding	0.26	7.60E-03
		Loss of Acetylation at K274	0.21	0.03
D269H	0.848	Altered Disordered interface	0.28	0.03
		Altered DNA binding	0.26	5.30E-03
		Gain of Acetylation at K264	0.24	0.02
		Loss of Ubiquitylation at K264	0.19	0.01
R254C	0.798	Altered Metal binding	0.68	1.80E-03
		Altered Disordered interface	0.5	9.30E-04
		Altered DNA binding	0.33	1.30E-03
		Gain of Relative solvent accessibility	0.24	0.04
		Loss of Acetylation at K256	0.19	0.05
R234W	0.657	Loss of Intrinsic disorder	0.53	4.80E-03
		Loss of Helix	0.32	2.50E-03
		Altered Disordered interface	0.29	0.03
		Loss of B-factor	0.26	0.04
		Gain of Acetylation at K236	0.19	0.05
		Altered DNA binding	0.14	0.05
		Altered Coiled coil	0.13	0.04

3.7. Predicting phenotypic effects of the nsSNPs

The analysis by Project HOPE to predict phenotypic effects for each of the mutations is shown in Table 5. The size and charge of the protein were impacted by each amino acid substitution. Five of these six locations had a charge change from positive to neutral, whereas one had a charge change from negative to neutral. These differences may affect how the protein interacts with other molecules. The hydrophobic residues of a protein normally remain together in the center, whereas the hydrophilic residues migrate to the surface, allowing the protein to become folded in the minimum-energy form. Four of the six mutations displayed a variation in hydrophobicity. HOPE was unable to produce structural pictures of both the changed amino acid residue at that specific protein location and the natural type because of the unavailability of a crystal structure. Furthermore, locations 271, 254, 234 and 315 are substantially preserved, according to conservancy analysis. As a result, this research also revealed that the protein would likely suffer from these mutations. While several additional residue types have also been detected at positions 269 and 464, these sites are likewise well conserved. The mutant amino acids did not belong to the types identified at this site in other homologous proteins. These two mutations may occasionally occur without harming the protein.

Table 5.

Prediction of the phenotypic effects of nsSNPs on TFEB by Project HOPE.

SNPs	Differ in size	Difference in charge	Difference in hydrophobicity	Disrupt hydrogen bond	Interferes with protein function	Interferes with other molecule/residue interaction	Domain/Region
R464C	YES	YES	YES	YES	NO	YES
R315H	YES	YES	NO	NO	YES	YES	leucine zipper
R271C	YES	YES	YES	YES	YES	YES	bHLH
D269H	YES	YES	NO	NO	YES	YES	bHLH
R254C	YES	YES	YES	YES	YES	YES	bHLH
R234W	YES	YES	YES	YES	NO	YES

3.8. Mutation cluster prediction

The amino acid residues at positions 254, 269 and 271 were discovered to belong to a cluster with a P value of 5.6e-3 when all 6 nsSNPs were submitted to mutation 3D web server which is shown in Fig. 3.

Fig. 3.

Mutation 3D for mutation cluster analysis. The location of the covered mutation (234) is indicated in blue, whereas the positions of the clustered mutations (254, 269 and 271) are indicated in red. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

3.9. TFEB-protein interaction prediction

According to the STRING service, 7 proteins interact with the TFEB protein (Fig. 4). These proteins include microphthalmia-associated transcription factor (MITF), tyrosine 3-monooxygenase/tryptophan 5-monooxygenase activation protein gamma (YWHAG), transcription factor E3 (TFE3), Ras-related GTP-binding protein C (RRAGA), Ras-related GTP-binding protein A (RRAGA) and mammalian target of rapamycin (mTOR). By using the maximum confidence (0.900), the interaction was predicted. The interactions have been determined empirically or through carefully selected databases. Gene cooccurrences, gene fusions, and gene neighborhoods are additional ways to predict some. Protein homology, coexpression, and text mining have been employed to determine protein homology. These interactions are indicated by different colored lines. These amino acid changes may also have an impact on the function of the interacting molecules since it was predicted that a number of the nsSNPs would disrupt the ability of TFEB to connect with other molecules.

3.10. Identification of nsSNPs in protein domains

The TFEB protein domain regions were predicted via the InterPro program. The primary domains include the finulus domain (241–294), microphthalmia-associated transcription factor N-terminus (4–162), and basic helix-loop-helix domain (226–316). The P values and the websites that predicted the domains are included in Table 6. Four out of the 6 deleterious nsSNPs were found in the helix-loop-helix DNA-binding domain (R271C, D269H, R254C, R234W).

Table 6.

Predicted domains of the TFEB by different websites.

Domain name	Start position	End position	P value	Prediction webserver
HLH, helix-loop-helix DNA-binding domain	229	298	3.66E-15	SUPERFAMILY
bHLHzip-TFEB	226	316	2.61E-62	CDD
MITF/TFEB/TFEC/TFE3 N-terminus	4	162	1.10E-56	Pfam
microphthalmia-associated transcription factor isoform X1	245	327	1.50E-45	FunFam
finulus	241	294	1.70E-15	SMART
Helix-loop-helix DNA-binding domain	236	289	1.90E-09	Pfam

3.11. Prediction of posttranslational modification sites (PTMs)

Using PhosphoSitePlus and iPTMnet, distinctive posttranslational alterations in TFEB were identified. Fig. 5 shows the PTM sites predicted by PhosphoSitePlus. Moreover, none of the chosen nsSNPs match any of the sites predicted by this server. iPTMnet indicates that R254 is the methylation site. Therefore, a change to cysteine affects the PTM activity of TFEB. Furthermore, the phosphorylation site, S463, is located immediately before R464C. iPTMnet indicates that none of the other mutation sites are PTM sites. The predictions by IPTMnet are given in Supplementary Table 5.

Fig. 5.

Prediction of posttranslational modification sites with the total number of references for PTM sites in the TFEB protein via PhosphoSitePlus.

Fig. 4.

TFEB protein‒protein interaction network with the STRING server.

3.12. 3D modeling of proteins

There was no crystal structure found for the entire 3-D structure of the TFEB protein. We employed the Alphafold Predicted Structure (AF-P19484-F1) as a model (https://www.uniprot.org/). Using the TM-Align service to align the Alphafold structure and crystal structure (PDB ID: 7Y62) and calculate the RMSD and TM score, the feasibility of this structure was verified (https://www.uniprot.org/). A sequence-independent structural comparison was carried out by TM-align, and the outcome represented by the TM score scales the topological similarity between two structures. A score of 1 denotes similarity between the wild-type and the query structure. The value ranges from 0 to 1. For Alphafold and 7Y62, the TM score is 0.67. The average difference between the corresponding atoms of two proteins is reflected in the RMSD (root-mean-square deviation) number, which was used to calculate the second score. A lower number signifies a greater degree of similarity between two structures. Alphafold and 7Y62 have an RMSD of 1.84.

SWISS-Model was used to predict the modified structures. Following the completion of the predictions, the protein model quality was checked via the QMEAN score, MolProbity score, and Ramachandran plot obtained from the SWISS-MODEL structure assessment tool, as were the ERRAT and PROCHECK scores obtained from the SAVES server. The results are given in Table 7. A protein structure's preferred Ramachandran region should ideally be greater than 98%. Prior to refining, the wild-type model (Alphafold) in PROCHECK was 74.4% in the Ramachandran favored zone and 13.1% of its residues were in the allowed region. Prior to refining, PROCHECK's Ramachandran favored region contained 74–75% of the residues in all of the projected models in this investigation of the mutant ones. With values close to 0, QMEAN Z scores represent the degree of nativeness of a specific protein structure. Low-quality models are indicated by scores below −4.0, with 0 denoting a structure similar to that of a native. Every model in this investigation had a Z score of approximately −9. The MolProbity score, a protein quality score that indicates the crystallographic resolution, was another assessment score. The lowest possible score is what is needed here. The ERRAT score is another instrument that is used to evaluate the integrity of protein structures. Higher scores for the ERRAT indicate higher quality, which is regarded as an overall quality factor for nonbonded atomic interactions. Additionally, PROCHECK, which gives the fraction of residues in the Ramachandran-favorable zone, was used.

Table 7.

Evaluation scores of the models before refinement.

Mutation	Template	ERRAT score	Ramachandran favored in PROCHECK	QMEAN	MolProbity Score	Tm score	RMSD
R464C	Alphafold	95	74.90%	−9.65	1.77	0.99999	0.03
R315H		94.5701	74.60%	−9.22	1.95	0.99997	0.04
R271C		94.5701	74.10%	−9.5	1.9	0.99998	0.03
D269H		94.5455	74.60%	−9.77	1.77	0.99999	0.03
R254C		94.5701	74.40%	−9.89	1.86	0.99999	0.03
R234W		94.2478	75.90%	−9.2	1.82	0.99992	0.07

3.13. Model refinement and evaluation

The alphafold and all of the altered models were refined via the GalaxyRefine web server. The Swiss model evaluation was utilized to assess the five models that are produced by the server for each. For the relevant SNPs, the model with the largest percentage of residues in the Ramachandran plot's most desired areas was selected. The projected model was assessed via the ERRAT and PROCHECK assessment tools. Table 8. shows the quality validation scores of the tools for each model after refinement. The evaluation tools revealed that each of the models was acceptable.

Table 8.

Evaluation scores of the models after refinement.

Mutation	ERRAT score	Ramachandran favored in PROCHECK	QMEAN	MolProbityScore	Tm score	RMSD
R464C	94.6502	94.00%	−0.39	0.84	0.9911	0.75
R315H	95.8678	94.20%	0.26	0.82	0.99172	0.72
R271C	96.2963	94.50%	0.17	0.77	0.9908	0.76
D269H	96.281	93.50%	0.03	0.89	0.99246	0.69
R254C	95.4545	94.70%	0.15	0.75	0.99122	0.75
R234W	95.082	95.50%	0.07	0.87	0.98986	0.8

3.14. Associations between high-risk nsSNPs and cancer susceptibility

The evaluation of high-risk nonsynonymous SNPs (nsSNPs) using multiple computational tools revealed significant associations with cancer susceptibility. CScape predicted all variants, except D269H, to be oncogenic, with R315H identified as a highly confident carcinogenic mutation. Dr. Cancer consistently classified all variants as disease-associated, including D269H, which was otherwise considered benign by CScape. The Cancer Genome Interpreter and FATHMM generally categorized most variants as passenger mutations; however, FATHMM specifically flagged R234W as cancer-related, suggesting a potential driver role. Among these, R254C gained further importance as it was detected in glioblastoma multiforme and uterine endometrioid carcinoma samples through cBioPortal analysis, providing clinical relevance. Overall, R315H, R254C, and R234W emerged as the most noteworthy variants, showing consistent oncogenic predictions and/or patient-level validation. All cancer associated results are presented in Table 9.

Table 9.

Prediction of high-risk nsSNPs linked to cancer susceptibility.

Variant ID	Mutation	CScape	Dr. cancer	Cancer genome interpreter	FATHMM (cancer)
rs369809425	R464C	Oncogenic	Disease	Passenger	Passenger
rs371625604	R315H	Oncogenic (high conf.)	Disease	Passenger	Passenger
rs1463113136	R271C	Oncogenic	Disease	Passenger	Passenger
rs1281647905	D269H	Benign	Disease	Passenger	Passenger
rs1561848484	R254C	Oncogenic	Disease	Passenger	Passenger
rs1327480600	R234W	Oncogenic	Disease	Passenger	CANCER

3.15. KEGG pathway analysis of TFEB

KEGG pathway analysis revealed that TFEB, functioning as transcription factors, are involved in the calcium signaling pathway (Supplementary Fig. 1). This pathway plays a pivotal role in regulating diverse cellular processes, including proliferation, differentiation, and stress responses. In addition, TFEB and NFAT were found to participate in apoptosis-related pathways, highlighting their contribution to cell survival and programmed cell death. Furthermore, their involvement in the MAPK signaling cascade suggests a broader regulatory role in controlling gene expression and cellular adaptation to extracellular stimuli (Supplementary Fig. 1). Together, these findings indicate that TFEB act as critical regulatory nodes linking calcium signaling to apoptosis and MAPK-mediated cellular responses.

3.16. Molecular docking analysis

The binding energy between the native and mutant TFEB proteins with a DNA E-box that contains destabilizing nonsynonymous mutations (R464C, R315H, R271C, D269H, R254C, and R234W) was investigated via HADDOCK. To comprehend how biological partners interact, the HADDOCK score needs to be calculated. During docking, all the structures were given a HADDOCK score, which helps to categorize the structures. The buried surface area (BSA), intermolecular AIR energy, van der Waals, electrostatic, and desolvation energies are used to calculate the score. A lower HADDOCK score indicates better interaction.⁷⁹ Given that Haddock's initial cluster is the best, the natural TFEB and destabilizing mutant TFEB bound with the DNA complex Haddock scores were obtained. The docking between TFEB and DNA is shown in Fig. 6. The interacting residues are taken from DNAproDB. The HADDOCK score was highest for the alphafold-DNA complex (−34.1 ± 1.4) and lowest for the R315H-DNA mutant structure (−97.1 ± 3.4). Furthermore, all the mutant-DNA complexes presented lower HADDOCK scores than did the wild-type alphafold-DNA complex. The native complex has a BSA value of 1727.8 ± 70.6, whereas the BSA values of the destabilizing nonsynonymous mutations of the TFEB–DNA complexes are between 1598.0 ± 143.5 and 2336.2 ± 100.3. The scores are displayed in Table 10.

Fig. 6.

Docking between TFEB and DNA. a-Alphafold-DNA, b-R234W-DNA, c-R254C-DNA, d-D269H-DNA, e-R271C-DNA, f-R315H-DNA, g-R464C-DNA.

Table 10.

Docking analysis of wild-type and mutant TFEB with DNA via HADDOCK.

Models	HADDOCK score	RMSD from the overall lowest-energy structure	van der Waals energy	Electrostatic energy	Desolvation energy	Restraints violation energy	Buried Surface Area	Z Score
Alphafold –DNA	−34.1+/-1.4	7.2 ± 0.2	−75.3 ± 6.0	−147.2+/-89.7	13.1 ± 6.9	575.5 ± 58.0	1727.8 ± 70.6	−1.2
R234W–DNA	−68.7 ± 6.9	7.0 ± 0.2	−80.6 ± 3.3	−266.7+/- 20.5	11.2 ± 1.6	540.9 ± 53.1	1734.6 ± 37.3	−1.5
R254C– DNA	−53.3+/- 11.9	6.2 ± 0.1	−72.4 ± 2.9	−168.5+/- 27.6	6.7 ± 2.8	461.7 ± 63.7	1610.3 ± 57.0	−1.7
D269H– DNA	−70.0+/- 12.4	1.3 ± 0.8	−84.3 ± 7.3	−231.5+/- 21.4	12.3 ± 3.1	483.3 ± 61.8	1810.3 ± 132.5	−1.5
R271C– DNA	−40.9 ± 6.0	4.6 ± 0.4	−73.2+/- 5.4	−155.0+/- 31.9	15.5 ± 3.1	478.5 ± 86.0	1598.0 ± 143.5	−1.2
R315H– DNA	−97.1 ± 3.4	1.1 ± 0.7	−109.1 ± 8.8	−268.9+/- 13.3	7.9 ± 3.2	579.5 ± 44.7	2336.2 ± 100.3	−2.3
R464C– DNA	−82.8 ± 6.4	2.1 ± 0.4	−90.5 ± 4.0	−252.6 ± 18.4	9.8 ± 2.6	485.0 ± 53.8	1916.7 ± 142.7	−1.9

DNAproDB was used to examine the interactions between proteins and DNA. Every protein was bound to the DNA molecule via hydrogen bonding and VdW interactions. The R271C mutant and DNA were shown to contain the most hydrogen bonds. DNA and the R234W mutant had the lowest number of hydrogen bonds. R315H had the most VdWs, whereas the alphafold had the lowest VdWs. The secondary structure composition was asymmetric, and the R315H mutant presented a good hydrophobicity score. The accessible surface area (BASA) of the buried solvent between specific residues and nucleotides was also obtained. Every score is shown in Table 11.

Table 11.

DNA‒protein Interfaces from DNAproDB.

Model	DNA entity ID	Pro. chain ID	Pro. chain segments	Nuc-Res interactions	Weak nuc-res interactions	Total BASA [Å2]	Total H-bonds	Total vdW	Hydrophobicity Score (SAP)	Secondary structure composition
Alphafold-DNA	B1@B2	A	A1	35	1	1058.152	6	90	−0.705	helix
R234W-DNA	B1@B2	A	A1	31	7	1076.549	5	102	−0.883	helix
R254C-DNA	B1@B2	A	A1	34	4	1149.3	6	98	−0.724	helix
D269H-DNA	B1@B2	A	A1	42	5	1333.606	7	113	−0.642	helix
R271C-DNA	B1@B2	A	A1	44	3	1116.268	14	99	−0.636	helix
R315H-DNA	B1@B2	A	A1	53	0	1451.863	9	120	0.288	irregular
R464C-DNA	B1@B2	A	A1	46	2	1400.092	9	115	−0.746	helix

The residues involved in the interaction between the TFEB alphafold and mutant proteins and DNA were analyzed via DNAproDB. The number of residues involved in the interaction between alphafold and DNA is 9, whereas in mutant proteins, it ranges from 8 to 18 residues. All the interacting residues are shown in Table 12. The interactions between the proteins and the DNA molecules are shown in Figs. 7A and 7B.

Table 12.

Positions of the residues that interact with the wild-type and mutant TFEB proteins and with DNA.

Model	DNAproDB interacting residues
alphafold DNA	136, 249, 256, 260, 262, 264, 265, 268, 272
R234WMP-DNA	244, 248, 251, 256, 269, 270, 271, 272, 273, 274, 275, 278
R254CMP-DNA	245, 249, 256, 260, 264, 268, 271, 272
D269HMP-DNA	249, 252, 256, 259, 260, 264, 268, 269, 270, 271, 272, 273, 274, 277
R271CMP-DNA	241, 248,268, 269, 270, 271, 272, 273, 274, 275
R315HMP-DNA	100, 106, 107, 108, 109, 112, 249, 252, 256, 259, 260, 264, 268, 270, 271, 272, 273, 274
R464CMP-DNA	91, 95, 98, 245, 249, 252, 256, 259, 264, 268, 269, 270, 271, 272, 273, 274, 277

Fig. 7A.

Interacting residues between proteins and DNA. a-Alphafold-DNA, b-R234W-DNA, c-R254C-DNA, d-D269H-DNA, e-R271C-DNA, f-R315H-DNA, g-R464C-DNA.

Fig. 7B.

The interacting residues between proteins and DNA. a-Alphafold-DNA, b-R234W-DNA, c-R254C-DNA, d-D269H-DNA, e-R271C-DNA, f-R315H-DNA, g-R464C-DNA.

4. Discussion

Transcription factors and enhancers are pivotal factors involved in disease and cancer progression. TFs undergo modifications in their functions across various diseases, either directly through mechanisms such as point mutations, translocations, amplifications, deletions, and altered expression levels or indirectly via processes that influence their binding to promoters. These alterations in TF activity can disrupt normal gene expression patterns, contributing to disease progression, including cancer. Understanding these intricate regulatory mechanisms is crucial for unraveling the molecular basis of diseases and for devising effective therapeutic interventions.80, 81, 82 A fundamental factor for the irregular expression of a protein is its SNPs. SNPs are common in the human genome and vary from individual to individual. They may alter the amino acids within the protein sequence. They are also present in the intronic portions of a gene. Studying the relationship between SNPs and their functional and structural impact on a protein helps to identify potential biomarkers for disease diagnosis and prognosis.

This study aimed to identify disease-causing nsSNPs in the TFEB. We combined sequence-based tools like SIFT and PROVEAN with machine-learning methods such as PolyPhen-2, as well as disease-association predictors like PhD-SNP and SNPs&GO, to screen and prioritize deleterious variants.⁸³ From the overall pool, six high-risk nsSNPs were selected based on their predicted impact on protein stability and potential functional disruption. To assess structural consequences, these variants were modeled using AlphaFold as template, following the structural analysis approach adopted in recent studies. Our methodology aligns closely with other published pipeline.83, 84, 85 Finally, given TFEB’s central role as a transcription factor influencing transcriptional regulation, autophagy, and lysosomal function, studying its nsSNPs is crucial for understanding pathogenic mechanisms. Similar investigations have demonstrated the importance of characterizing transcription factor variants to elucidate functional outcomes.⁸⁶

TFEB belongs to the microphthalmia (MiT) family, which is a basic helix-loop-helix (bHLH)-leucine zipper TF. The other members include microphthalmia-associated TF (MITF), TFE3 and TFEC.² It regulates lysosomal and auto-phagosomal biogenesis and helps cells adapt to stress during starvation and energy depletion. Recent studies have also shown increased regulatory activities of TFEB (metabolism, immunity, angiogenesis and inflammation).⁹ TFEB regulates lysosomal gene promoters, which are known as coordinated lysosomal expression and regulation motifs (GTCACGTGAC; CLEAR). It is a palindromic consensus sequence that overlaps the E-Box.⁷ TFEB orchestrates autophagy and lysosome functions and is a potential therapeutic target in lysosome storage diseases.19, 87

The protein–protein interaction network analysis of TFEB by STRING highlights its interactions with other proteins underscoring the potential for nsSNPs to disrupt these associations and contribute to pathological outcomes.⁸⁸ For instance, TFEB's binding to 14-3-3/YWHA proteins, which sequesters it in the cytoplasm under nutrient-rich conditions, is mediated by phosphorylation sites; mutations in the interaction interface disrupt binding, leading to altered expression of target and dysregulated autophagic flux.⁸⁹ Moreover, MITF, TFE3 and TFEB have complex interactions in melanoma cells.⁹⁰ It was found in previous studies that these interacting proteins which are functionally kinases, signaling adaptors, or transcriptional co-regulators, play critical roles in TFEB's function.89, 90 This helps reveal how nsSNP-induced changes could impair TFEB's phosphorylation-dependent localization, transcriptional activation, or pathway coordination, offering mechanistic insights into associated disorders where TFEB dysfunction is implicated.⁹¹

Domain analysis revealed that 4 out of the 6 deleterious nsSNPs were located in the helix-loop-helix DNA-binding domain (R271C, D269H, R254C, R234W). The mutation 3D web server revealed that the amino acids at positions 254, 269 and 271 belong to a cluster. As the mutations are in the helix-loop-helix DNA binding domain of the TFEB gene, those mutations might affect the function of the TFEB protein and can cause potential diseases. The basic region of the bHLH interacts directly with the major groove of DNA, specifically recognizing palindromic E-box motifs in the promoters of target genes involved in autophagy and lysosomal biogenesis.² Functionally, the bHLH domain's integrity is vital for TFEB's role in autophagy-lysosomal pathway (ALP) regulation, as mutations here can lead to cytoplasmic mislocalization or reduced nuclear activity.⁸⁹

When TFEB is phosphorylated at the 142nd (Ser142) and 211st (Ser211) sites together, it remains in an inactive state. For its activation and transportation into the nucleus, it needs to be dephosphorylated at either Ser142 or Ser211 for its regulatory function.12, 13, 92 The mammalian target of rapamycin (mTOR) serine/threonine kinase is the main protein involved in TFEB phosphorylation.⁹² Some studies also suggest that other kinases, such as extracellular regulated protein kinase 2 (ERK2)⁸ and protein kinase C (PKC),⁹³ may also be involved in TFEB phosphorylation. The regulatory role of mTOR in TFEB nuclear localization is more complex. The substitution of S462, S463, S466, S467 and S469 with phosphomimetic aspartate residues causes TFEB to translocate into the nucleus.⁹⁴ In this study, R464 was also found to be associated with carcinogenesis. This might be caused by the structural disturbance in the S462--S469 region.

The CBioPortal server revealed that R315C is associated with cutaneous melanoma and that R315S and R271H are associated with uterine endometrioid carcinoma. In our study, we predicted R315H and R271C to be associated with cancer through our in silico tools. R315H was also found to be associated with cancer with high confidence through the CScape server. These two SNP points might be important for TFEB structure and functions where different changes in amino acids can occur and may cause different types of diseases and cancers. R315H had the highest HADDOCK score, which indicates that this mutant TFEB binds more efficiently with DNA. The number of interacting residues in the R315H-DNA complex was also greater than that in the other complexes. The HADDOCK score for all the mutant and DNA complexes was lower than that for the alphafold DNA complex.

TFEB is integrated into several nutrient-sensing and stress-response pathways, where its phosphorylation status dictates its function. During starvation or lysosomal stress, mTORC1 inhibition leads to TFEB dephosphorylation by calcineurin (activated via lysosomal calcium release through MCOLN1), enabling nuclear entry and activation of autophagy-lysosomal gene.⁹⁵ AMP-activated protein kinase (AMPK) phosphorylates TFEB enhancing its transcriptional activity independently of localization, often by inhibiting mTORC1.⁹⁶ In the Wnt pathway, TFEB undergoes PARsylation, forming a complex with β-catenin-TCF/LEF1 to regulate specific target genes distinct from canonical Wnt targets.⁹⁵ TFEB overexpression may occur as a result of mutations. Lysosomal activity as well as lysosomal enzyme levels and lysosomal numbers are increased by TFEB overexpression.⁶ It has been demonstrated that pancreatic adenoductal carcinoma cells overexpress TFE3, TFEB, and MITF, indicating that they stimulate autophagy to further tumor growth.²⁵ RagD is directly controlled by TFEB at the transcriptional level. Tumors with elevated levels of MiT-TFE genes, such as TFEB, presented increased RagD expression. This elevated RagD expression leads to increased activity of mTORC1, a signaling pathway associated with cell growth and metabolism regulation.²⁶ A study also revealed correlations between functional SNPs of TFEB and multiple neurodegenerative illnesses, such as Huntington's disease, Parkinson's disease and Alzheimer's disease.²⁷ Moreover, High expression of TFEB is associated with many cancers such as colorectal cancer,⁹⁷ breast cancer,⁹⁸ non-small cell lung cancer.⁹⁹

One major limitation of this study is its exclusive reliance on computational predictions without experimental validation or systematic prioritization of candidate variants. As highlighted in previous reports, in silico analyses may not fully capture biological complexity and therefore remain hypothesis-generating in nature.¹⁰⁰ Furthermore, although DNA docking was prioritized due to TFEB’s role as a transcription factor, this approach does not fully capture the potential impact of nsSNPs on TFEB function. TFEB activity is also modulated through critical protein–protein interactions, which were not examined in the present study. As noted in prior computational investigations, limiting structural analysis to a single interaction context may restrict biological interpretation.¹⁰⁰

In conclusion, our study identified 6 nsSNPs in the TFEB gene which negatively affect the structural and functional properties of the TFEB protein. A strong correlation between cancer and mutations in R315 of TFEB gene has also been identified. This comprehensive analysis can inform future research on TFEB, enabling the exploration of potential disease-causing SNPs and aiding in the identification of effective drugs or pharmacological targets. Therefore, further experimental mutational research, genome-wide association studies, and clinical-based studies are necessary to validate these findings.

Data availability

All data generated or analysed during this study are included in this manuscript [and its supplementary material files].

CRediT authorship contribution statement

Mohtasim Fuad: Writing – original draft, Software, Methodology, Formal analysis, Data curation. Sadia Akter: Writing – original draft, Software, Methodology, Formal analysis, Data curation. Zimam Mahmud: Writing – review & editing, Writing – original draft, Validation, Supervision, Resources, Project administration, Methodology, Investigation, Funding acquisition, Formal analysis, Conceptualization. Sonia Tamanna: Writing – review & editing, Validation, Resources, Investigation, Formal analysis. Mohammad Sayem: Writing – review & editing, Validation, Software, Methodology, Investigation, Formal analysis. ABM Reazul Islam Zim: Writing – review & editing, Validation, Software, Methodology, Formal analysis. Md. Zakir Hossain Howlader: Writing – review & editing, Visualization, Validation, Software, Resources, Investigation, Formal analysis.

Funding

This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Appendix A. Supplementary data

The following are the Supplementary data to this article: