Bioinformatics-driven identification of pathogenic missense nsSNPs in the human proto-oncogene SRC and cancer susceptibility

Highlights

• •Missense SNPs W151C, Y419N, and P465S in SRC disrupt protein structure and function.
• •Molecular dynamics simulations showed variants alter protein stability and flexibility.
• •Docking analysis reveal variants affect ligand binding affinity.
• •SRC mutations influence the expression of several genes in multiple myeloma and uterine cancer.

Abstract

SRC is a proto-oncogene that regulates cell proliferation and survival, and its dysregulation is commonly observed in diverse cancers. While SRC kinase dysregulation is well-established as a cancer driver, the functional consequences of its genetic variants, particularly non-synonymous single-nucleotide polymorphisms (nsSNPs) are not fully understood. Therefore, we employed an integrative computational approach to identify nsSNPs in SRC and analyze their impact on protein function and structure. Out of the 512 missense nsSNPs analyzed, 42 were predicted to be deleterious, with 12 likely to destabilize protein structure. Among these, three mutations, namely W151C (rs746439256), Y419N (rs2147125119), and P465S (rs1251532695), were particularly significant, causing substantial physicochemical changes. Molecular dynamics simulations revealed that these variations reduce protein stability and flexibility, resulting in conformational alterations. Docking study demonstrated that these mutations disrupt the binding interface residues of the SRC-FAK complex and affect dasatinib binding affinity. Additionally, gene expression analysis linked mutated SRC to dysregulation of cancer-related genes, especially in multiple myeloma and uterine cancer, and suggested reciprocal regulation by other mutated genes across malignancies. These findings highlight the oncogenic potential of SRC mutations and pave the way for future population-based studies exploring their role as diagnostic biomarkers, therapeutic targets, and modulators of drug response in personalized cancer treatment.

1. Introduction

The human SRC gene was first identified as the transforming agent of the avian Rous sarcoma virus, which led to the discovery of its cellular homolog, c-SRC, a proto-oncogene that encodes a non-receptor tyrosine kinase.¹ SRC is a member of the SRC family of kinases (SFKs), which also includes FYN, HCK, YRK, LYN, YES, FGR, BLK, and LCK, all of which play critical roles in cellular signaling networks.² As a non-receptor tyrosine kinase, SRC functions through protein–protein interactions rather than direct ligand binding, regulating key cellular processes such as adhesion, motility, cell cycle progression, immune responses, and apoptosis.3, 4, 5

Dysregulation of SRC through overexpression or aberrant kinase activity has been implicated in a wide range of malignancies, including glioblastoma, melanoma, lung, breast, colon, liver, prostate, pancreatic, bladder, hematological, and head and neck cancers.6, 7, 8, 9 Notably, SRC is overexpressed in approximately 80 % of colorectal cancer cases and exhibits increased activity at metastatic sites, correlating with poorer outcomes and reduced disease-free survival.10, 11 Moreover, a truncating mutation at codon 531 has been detected in 12 % of advanced cases, leading to constitutive SRC activation that drives tumorigenesis and promotes metastatic spread.¹² Similarly, in non-small cell lung cancer, SRC activation enhances angiogenesis and tumor growth, acting synergistically with the MEK and STAT3 signaling pathways.13, 14

Single-nucleotide polymorphisms (SNPs) are the most common type of genetic variation and are defined as single-base substitutions occurring at a frequency of at least 1 % in the human population. Although they comprise approximately 0.1 % of the genome, SNPs are widely distributed across both coding and non-coding regions.15, 16, 17 Within coding regions, missense non-synonymous SNPs (nsSNPs) are particularly significant because they can alter amino acid sequences and disrupt protein structure or function.¹⁸ These changes may impair SRC kinase activity, thereby influencing downstream signaling pathways. For example, a germline gain-of-function SRC mutation (p.E527K) has been identified as the cause of inherited thrombocytopenia syndrome, wherein constitutive SRC activation impairs platelet production.19, 20 This highlights the significant impact that a single amino acid substitution in the SRC protein can have on its normal physiological function.

Given the oncogenic role of SRC, identifying pathogenic nsSNPs is crucial for understanding variant-specific mechanisms underlying tumorigenesis and advancing the development of more precise, genotype-informed therapeutic strategies. Unlike traditional experimental approaches, which are labor-intensive, time-consuming, and costly, in silico methods, such as protein structural modeling, molecular docking, molecular dynamics simulations, and pathogenicity prediction algorithms, offer efficient and cost-effective strategies to streamline and expedite the identification of deleterious variants.²¹ These techniques also provide valuable insights into variant-induced perturbations in SRC protein structure and function.

Despite the well-established oncogenic role of SRC, no comprehensive computational study has systematically analyzed the structural and functional consequences of its missense variants. This limits our understanding of how specific SRC nsSNPs contribute to cancer susceptibility and progression. To address this, the present study employed a multi-faceted in silico pipeline to identify promising deleterious SRC variants, characterize their structural and functional consequences, and predict their effects on protein–protein and protein–ligand interactions. This work provides a systematic computational framework for prioritizing SRC nsSNPs for experimental validation in vitro and in vivo, thereby advancing our understanding of SRC-mediated oncogenesis and supporting targeted therapeutic development.

2. Materials and methods

2.1. SNP data retrieval

SNP data for the SRC gene (ID: 6714) were sourced from the NCBI dbSNP database (https://www.ncbi.nlm.nih.gov/snp/).²² A total of 512 missense SNPs (nsSNPs) were retrieved because of their potential to alter amino acids in proteins, which can lead to abnormal function and disease phenotypes. For each nsSNP, the accumulated data comprised the reference SNP ID (rsID), alleles, chromosomal location, protein accession number (NP_938033.1), and corresponding amino acid changes. The nucleotide locations of these missense nsSNPs were mapped to the Human Genome Assembly GRCh38.p14. Data collection was completed on January 7, 2024. The study methodology is shown in Fig. 1.

Fig. 1.

Flowchart outlining methodological steps used in the study.

2.2. Prediction of potential high-risk missense nsSNPs

Nine computational tools were used to assess the functional consequences and pathogenicity of missense SNPs, and those predicted as high-risk by all nine tools were selected using a consensus method. We employed three tools to identify deleterious nsSNPs. PredictSNP1 (https://loschmidt.chemi.muni.cz/predictsnp1/) is a consensus classifier that integrates eight tools to predict the impact of nsSNPs.²³ Sorting Intolerant From Tolerant (SIFT; https://sift.bii.a-star.edu.sg/) uses sequence homology and physical characteristics to predict the implications of amino acid substitutions, considering those with a probability score $<$ 0.05 as deleterious.²⁴ The Local Identity and Shared Taxa (LIST; https://gsponerlab.msl.ubc.ca/software/list/) estimates the impact of nsSNPs on protein function in closely related species, considering scores $\ge$ 0.85 as deleterious.²⁵

In the next step, damaging nsSNPs were analyzed using three web-based tools. Polymorphism Phenotyping v2 (PolyPhen2; http://genetics.bwh.harvard.edu/pph2/) is another tool that uses sequence, phylogenetic, and structural features, categorizing SNPs as “probably damaging” ($>$ 0.9), “possibly damaging” (0.4–0.9), or benign ($\le$ 0.4).²⁶ Protein Analysis Through Evolutionary Relationship (PANTHER; https://www.pantherdb.org/) classifies proteins based on evolutionary relationships and molecular functions.²⁷ Functional Analysis through Hidden Markov Models (FATHMM; https://fathmm.biocompute.org.uk/) predicts the impact of nsSNPs on protein structure and function using Hidden Markov Models.²⁸

Lastly, three additional tools are employed to predict pathogenic missense SNPs. SNPs&GO (https://snps.biofold.org/snps-and-go/snps-and-go.html) uses support vector machines (SVM) to predict disease-causing nsSNPs.²⁹ Meta-SNP (https://snps.biofold.org/meta-snp/) is a consensus classifier integrating four tools to predict disease-associated nsSNPs.³⁰ MutPred2 (http://mutpred.mutdb.org/) is a machine learning-based tool predicting the pathogenicity of amino acid substitutions.³¹ Mutations with scores $>$ 0.5 are estimated to be pathogenic by these tools.

2.3. Predicting the impact of missense nsSNPs on protein stability

Stability perturbations of the SRC protein due to nsSNPs are analyzed using seven different in silico tools, which were categorized into two groups. The first group included sequence-based tools such as iStable (http://predictor.nchu.edu.tw/iStable/),³² MUpro (https://mupro.proteomics.ics.uci.edu/),³³ and I-Mutant2 (https://folding.biofold.org/i-mutant/i-mutant2.0.html).³⁴ These tools predict the impact of amino acid substitutions on protein stability based on the protein sequence at 25 °C and pH 7.

The second group consisted of structure-based tools, including SDM (http:// www-cryst.bioc.cam.ac.uk/∼sdm/sdm.php),³⁵ CUPSAT (https://cupsat. brenda-enzymes.org/),³⁶ DynaMut (https://biosig.lab.uq.edu.au/dynamut/), and DUET (https://biosig.lab.uq.edu.au/duet/).³⁷ These tools utilized the PDB structure of the wild-type SRC protein, along with nsSNPs. The impact of nsSNPs on SRC protein stability is determined using Gibbs free energy change ($\mathrm{\Delta }\mathrm{\Delta }$ G) in kcal/mol and a confidence score. A $\mathrm{\Delta }\mathrm{\Delta }$ G and confidence score below 0 indicates decreased stability, whereas a score above 0 suggests increased stability.

2.4. Evolutionary conservation and phylogenetic relationship analysis

Evolutionary conservation analysis was performed using ConSurf, which applies multiple sequence alignments and Bayesian inference to assign residue-level conservation scores (1–4: variable; 5–6: intermediate; 7–9: highly conserved).³⁸ For this analysis, the SRC protein sequence was aligned against the UNIREF-90 protein database using the HMMER algorithm (E-value cutoff = 0.0001) and MAFFT-L-INS-i for sequence alignment. This approach facilitated the identification of conserved residues that are likely to be structurally and functionally significant.

Similarly, phylogenetic relationship analysis is another approach that enables the identification of conserved regions within protein sequences across diverse species. To conduct this analysis, homologous sequences of the human SRC protein were retrieved from various organisms using NCBI BLASTp (https://blast.ncbi.nlm.nih.gov/Blast.cgi?PAGE=Proteins).³⁹ Subsequently, these sequences underwent multiple sequence alignment (MSA) using Clustal Omega (https://www.ebi.ac.uk/jdispatcher/msa/clustalo),⁴⁰ facilitating the detection of conserved residues based on evolutionary patterns. The aligned sequences were visualized using ESPript3 (https://espript.ibcp.fr/ESPript/ESPript/),⁴¹ which provides a graphical representation of sequence conservation alongside secondary structure elements. A phylogenetic tree was also constructed using the Neighbor-Joining (NJ) method in MEGA software to demonstrate the evolutionary relationships among SRC homologs.⁴²

2.5. 3D structure modeling and validation

The homology modeling tool SWISS-MODEL (https://swissmodel.expasy.org/) was used to construct the three-dimensional (3D) structures of the wild-type and mutant SRC proteins.⁴³ Missense nsSNP-incorporated FASTA sequences of SRC were used to generate the 3D structures of the mutant models. The “Search For Templates” option was chosen to enable a more ordered and logical approach for template selection, rather than the default “Build Model” option, which automatically selects a template and starts the model generation. The appropriate template was selected based on a high identity score and a Global Model Quality Estimation (GMQE) value ranging from 0 to 1, with higher numbers indicating good model quality. SWISS-MODEL provides a Qualitative Model Energy Analysis (QMEAN) Z-score, where a score below −4.0 indicates a low-quality and unreliable model.⁴⁴ Mutant models were superimposed on the wild type for visualization and comparison using UCSF Chimera version 1.17.3.⁴⁵

Subsequently, the SRC models were validated using tools available on the SAVES v6.0 (https://saves.mbi.ucla.edu/) server, including ERRAT, Verify3D, and PROCHECK. ERRAT examines nonbonded atom–atom interactions, with high-quality structures scoring between 91 % and 95 %, whereas Verify3D evaluates the compatibility of the 3D structure with its amino acid sequence, with scores of more than 80 % suggesting a high-quality model.46, 47 PROCHECK was used to assess the stereochemical quality of the model, with a model amino acid residue percentage of more than 90 % in the most favored regions, indicating a high-quality model.⁴⁸ MolProbity (http://molprobity.biochem.duke.edu/) was used to analyze all-atom interactions and geometry, and the ProSA (https://prosa.services.came.sbg.ac.at/prosa.php) server was used to compare protein structures to X-ray and NMR-determined structures.49, 50

2.6. Determination of domain-specific missense nsSNPs and structural consequences

Domain analysis is crucial because it identifies functionally important regions of a protein, such as active and binding sites. The InterPro (https://www.ebi.ac.uk/interpro/) database, which classifies proteins into families and predicts conserved domains and functional sites, was used to determine the domains of the SRC kinase protein.⁵¹ Subsequently, the HOPE and Missense3D tools were used to investigate the influence of missense SNPs on the SRC protein structure. HOPE (https://www3.cmbi.umcn.nl/hope/) is a web-based tool that combines sequence, structural, and evolutionary data to assess the impact of amino acid alterations on protein function and structure.⁵² In parallel, Missense3D (https://missense3d.bc.ic.ac.uk/) predicts the potential of missense variants to disrupt protein structure by detecting alterations in structural properties, such as steric clashes, cavity formation, and hydrogen bonding.⁵³

Wild-type and mutant models of the SRC protein were prepared by removing the co-crystallized ligand using BIOVIA Discovery Studio 2025.⁵⁴ Polar hydrogens were added to the structures and subjected to energy minimization using the Swiss-PdbViewer.⁵⁵ These models were then applied to molecular dynamics simulations and docking analyses to understand the effect of missense SNPs on the dynamic structure and binding affinity of SRC protein.

2.7. Molecular dynamics (MD) simulations analysis

The detrimental effects of nsSNPs on the SRC protein structure were examined using 200-nanosecond (ns) MD simulations using the Desmond software package (Schrödinger LLC).⁵⁶ The simulations employed the OPLS_2005 force field and SPC water model.⁵⁷ To prepare the simulation system, the SRC protein was placed in an orthorhombic periodic boundary box, ensuring a 10 Å buffer distance from the protein center to maintain consistent volume. The temperature and pressure were set to 300 K and 1 atm, respectively, with a salt concentration of 0.15 M sodium chloride, neutralized with Na⁺ and Cl^- ions. Preliminary relaxation procedures, such as energy minimization in the NPT ensemble, were performed to ensure system stability before the MD simulations.

After the simulations, the trajectories were analyzed to identify the structural changes induced by the nsSNP. Structural properties, including root mean square deviation (RMSD), root mean square fluctuation (RMSF), radius of gyration (Rg), and secondary structure element (SSE), were analyzed to understand the mutational effects on the stability, flexibility, compactness, and conformational dynamics of the protein. Additionally, Principal Component Analysis (PCA) and Dynamic Cross-Correlation Matrix (DCCM) were used to assess the overall impact of missense SNPs on protein structure. PCA was used to capture the principal modes of motion, whereas DCCM identified changes in the correlated and anti-correlated movements of C $\mathrm{\alpha }$ atoms, indicating potential alterations in inter-residue communication. Analyses were performed using the Bio3D package in R.⁵⁸

2.8. SRC gene-gene interaction network analysis

The intricate network of interactions involving the SRC gene and its associated proteins and genes were studied using GeneMANIA and STRING (accessed August 18, 2024). GeneMANIA (https://genemania.org/) predicts gene-gene interactions by integrating genomic and proteomic datasets. This platform leverages multiple interaction types, including physical interactions, co-expression, genetic associations and biological pathways.⁵⁹ In contrast, STRING (vs. 12) (https://string-db.org/) was used to investigate protein–protein interaction networks through a multi-layered approach, incorporating data from neighborhood relationships, gene fusion events, co-occurrence, co-expression patterns, experimental evidence, curated database entries, text mining, and homology.⁶⁰ Functional partner genes in STRING are assigned a confidence score ranging from 0 to 1, where 0 indicates a low probability of interaction and 1 indicates a high likelihood of interaction.

2.9. Molecular docking analysis

Molecular docking was performed to assess the impact of nsSNPs on protein–protein and protein-drug binding affinity. In protein–protein docking analysis, SRC was docked with FAK because SRC activates FAK, which then phosphorylates the cytoskeletal adaptor protein paxillin. The SRC/FAK/paxillin complex regulates cell migration and motility, acting as a crucial mediator of the integrin signaling pathway.⁶¹ The 3D structure of FAK (PDB ID: 2ETM, Chain B) was obtained from the Protein Data Bank (PDB), and non-standard elements, including additional chains, water molecules, and heteroatoms, were removed to prepare the structure for docking.⁶² Docking was performed using ClusPro 2.0 (https://cluspro.bu.edu/login.php), which predicts the best binding poses by ranking conformations according to energy stability and using a fast Fourier transform (FFT) algorithm.⁶³ Previously prepared SRC wild-type and mutant PDB structures were used for docking, and docked complexes with the lowest energy scores and highest cluster members were selected because they indicated favorable and stable binding between the receptor and ligand. Subsequently, mCSM-PPI2, a machine learning-based tool, was used to assess the impact of substitution on protein–protein binding affinity.⁶⁴ Specific changes in the bonds and interacting residues within the SRC-FAK complex were identified using PDBsum (https://www.ebi.ac.uk/thornton-srv/databases/pdbsum/).⁶⁵

To investigate the impact of missense SNPs on drug-binding affinity, the tyrosine kinase inhibitor dasatinib (PubChem CID: 3062316) was docked with both wild-type and mutant forms of SRC. The chemical structure of dasatinib was retrieved in SDF format from PubChem.⁶⁶ Blind docking simulations were carried out using AutoDock Vina via PyRx.⁶⁷ Prior to docking, ligand structures were energy-minimized employing the Universal Force Field (UFF) with the Conjugate Gradients optimization algorithm. The docking grid was defined by specific center coordinates and dimensions, and the exhaustiveness parameter was set to eight for both SRC wild-type and mutant models. The best docking pose was determined by the lowest binding energy (kcal/mol) and RMSD values. The generated SRC-dasatinib complexes were visualized in Discovery Studio.

2.10. Prediction of malignant missense nsSNPs

Cancer-associated nsSNPs within the SRC gene were predicted using four tools, including CScape, Dr. Cancer, and CancerVar. These tools employ statistical models, machine learning, and deep learning to predict whether a mutation is likely cancer-associated or benign. Dr. Cancer (https://snps.biofold.org/drcancer/drcancer.html) applies machine learning to classify mutations, assigning a probability score above 0.5 as a disease association.⁶⁸ Based on Oncogenic Prioritization by Artificial Intelligence (OPAI) scores, CancerVar (https://cancervar.wglab.org/) leverages deep learning to categorize somatic mutations as oncogenic, benign, or unknown.⁶⁹ CScape (https://cscape.biocompute.org.uk/) predicts whether mutations are neutral or cancer-driving by assigning a p-value from 0 to 1, with values above 0.5 considered significant.⁷⁰ The input was formatted as chromosome, position, reference nucleotide, and mutant nucleotide.

2.11. Determining the regulatory impact of SRC and other gene mutations in cancer

To examine the regulatory effects of SRC and other gene mutations in various human malignancies, we used muTarget (https://www.mutarget.com/). This tool allows assessment of the impact of the mutated SRC gene on the expression of other genes, as well as the influence of mutations in other genes on SRC expression, by integrating genomic and transcriptomic data from The Cancer Genome Atlas (TCGA) database.⁷¹ Genes of interest were identified based on a Mann-Whitney U test p-value of $\le$ 0.05, a mean fold change (FC) $>$ 1.44, and a mutation prevalence of at least 2 %.

3. Results

3.1. Retrieval of missense nsSNPs in the SRC gene

SRC-related SNPs were retrieved from the NCBI dbSNP database, yielding 24,260 variants. These variations were grouped according to their genomic positions and functional significance. The majority of the variants (21,634; 77 %) were intronic, with those found in the 5′ untranslated region (UTR) (857; 3 %), 3′ UTR (935; 3 %), synonymous coding regions (299; 1 %), non-coding transcript regions (443; 1 %), regions within 500 base pairs downstream (182; 1 %), and 2 kilobase pairs upstream (1,299; 5 %). Intronic SNPs were abundant because introns constitute a considerably large portion of the SRC gene sequence. Another 1,999 variants (7 %) were classified into other categories, such as splice-acceptor (5), splice-donor (12), nonsense (13), and unspecified (1969) variants. Missense variations represented 512 (2 %) of all SNPs. Unlike synonymous or intronic variants, missense SNPs can affect protein structure and function; hence, these SNPs were prioritized. The distribution of different SNP types within the SRC gene, as detailed in the dbSNP database, is shown in Fig. 2A.

Fig. 2.

Distribution and filtration of high-risk SNPs in the human SRC gene. (A) Pie chart showing the percentage distribution of SNPs. (B) Number of deleterious missense SNPs. (C) Number of damaging missense SNPs. (D) Number of pathogenic missense SNPs.

3.2. Prediction of disease-associated missense nsSNPs

Missense SNPs of SRC were evaluated using nine sequence-based online tools to predict deleterious, damaging, or disease-associated/pathogenic SNPs. From the initial set of 512 nsSNPs, 482 remained after the processing. Among these, SIFT identified 135, PredictSNP1 detected 71, and LIST classified 232 cases as deleterious (Fig. 2B). For damaging SNPs, PolyPhen-2, PANTHER, and FATHMM predicted 339, 429, and 153 SNPs, respectively (Fig. 2C). PolyPhen-2 identified 84 possibly damaging, 255 probably damaging, and 143 benign SNPs. For pathogenic SNP classification, SNPs&GO, Meta-SNP, and MutPred2 predicted 178, 216, and 308 nsSNPs, respectively (Fig. 2D). From the 512 analyzed missense nsSNPs, 42 were filtered as high-risk based on a consensus across nine prediction tools and were carried forward for further analysis. The details of all analyzed missense SNPs are provided in Supplementary Table S1.

3.3. Stability assessment of pathogenic missense nsSNPs

To assess the potential structural and functional consequences of pathogenic nsSNPs, we evaluate 42 variants for their impact on protein stability. Protein stability was evaluated through predicted changes in Gibbs free energy (ΔΔG), a widely accepted metric for estimating the effect of amino acid substitutions on protein conformation. A lower ΔΔG score indicates destabilization of the protein structure. Seven previously mentioned web-based tools were used to assess changes in stability. Sequence-based tools such as I-Mutant2, MUpro, and iStable predicted 39, 40, and 38 missense SNPs, respectively, to be destabilizing. In comparison, structure-based tools, including SDM, DUET, CUPSAT, and Dyna-Mut, identified 34, 37, 26, and 23 SNPs as contributors to stability reduction (Supplementary Table S2). By integrating the outcomes from all seven tools, 12 nsSNPs were identified as destabilizing structures based on predictive consensus scores, as shown in Table 1.

Table 1.

List of 12 nsSNPs altering the stability of SRC protein.

SNP ID	Variants	I-Mutant2	MUpro	iStable	SDM	DUET	CUPSAT	Dyna-Mut
rs746439256	W151C	Decrease	Decrease	Decrease	Destabilizing	Destabilizing	Destabilizing	Destabilizing
rs2070646412	K203E	Decrease	Decrease	Decrease	Destabilizing	Destabilizing	Destabilizing	Destabilizing
rs2147108100	G277D	Decrease	Decrease	Decrease	Destabilizing	Destabilizing	Destabilizing	Destabilizing
rs994208712	A374T	Decrease	Decrease	Decrease	Destabilizing	Destabilizing	Destabilizing	Destabilizing
rs2147124810	V386D	Decrease	Decrease	Decrease	Destabilizing	Destabilizing	Destabilizing	Destabilizing
rs2147124842	R388Q	Decrease	Decrease	Decrease	Destabilizing	Destabilizing	Destabilizing	Destabilizing
rs2147125119	Y419N	Decrease	Decrease	Decrease	Destabilizing	Destabilizing	Destabilizing	Destabilizing
rs2147126336	G440D	Decrease	Decrease	Decrease	Destabilizing	Destabilizing	Destabilizing	Destabilizing
rs2147126445	I444N	Decrease	Decrease	Decrease	Destabilizing	Destabilizing	Destabilizing	Destabilizing
rs1251532695	P465S	Decrease	Decrease	Decrease	Destabilizing	Destabilizing	Destabilizing	Destabilizing
rs753395244	R483Q	Decrease	Decrease	Decrease	Destabilizing	Destabilizing	Destabilizing	Destabilizing
rs2147128460	M484V	Decrease	Decrease	Decrease	Destabilizing	Destabilizing	Destabilizing	Destabilizing

3.4. Evolutionary conservation and phylogenetic relationship analysis

The evolutionary conservation of amino acid residues in the native SRC protein was assessed using the ConSurf web server. This analysis identified structural and functional residues among the 12 high-risk nsSNPs of SRC, based on evolutionary conservation and solvent accessibility. Of these, residue W151 is classified as buried and structural, whereas residues P465, G277, R388, Y419, and R483 are exposed and functional. All six of these residues are highly conserved with conservation score of 9, except Y419 which has a conservation score of 8, as shown in Fig. 3. In addition, K203 is predicted to be conserved and exposed, while A374, V386, G440, I444, and M484 are conserved and buried. Functionally, exposed residues are likely to participate in molecular interactions, whereas buried residues contribute to the structural stability of the protein. Consequently, K203, A374, V386, G440, I444, and M484, with conservation scores of 7, 7, 8, 5, 7, and 8, respectively, were excluded, as these positions lack predicted functional or structural significance.

Fig. 3.

ConSurf prediction of conserved residue positions in the SRC protein affected by nsSNPs. Deleterious mutations are marked with red boxes. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

The analysis of sequence conservation among SRC orthologs revealed that several amino acid residues, including those corresponding to the 6 identified nsSNPs, are also found to be conserved across closely related primate species. Protein sequences from Homo sapiens (NP_938033.1), Pan troglodytes (XP_016793386.1), Macaca mulatta (NP_001248263.1), Pongo abelii (PNJ57733.1), Callithrix jacchus (XP_017827520.1), Piliocolobus tephrosceles (XP_023083713.1), Cebus imitator (XP_017367806.1), and Hylobates moloch (XP_031998312.1), were retrieved from NCBI BLASTp. These sequences were aligned using Clustal Omega, and the alignment was visualized with ESPript3 (Supplementary Fig. S1). All six nsSNPs positions were found within regions exhibiting high sequence conservation across the aligned species. A phylogenetic tree was constructed to represent evolutionary relationships among the selected primates, supporting the close genetic relatedness of these species (Fig. 4).

Fig. 4.

Phylogenetic tree of SRC protein of Homo sapiens with other closely related primate species.

3.5. 3D structure modeling and validation

3D modeling enables predicting SNP-induced structural changes by analyzing alterations of resultant proteins. The 2h8h.1.A was chosen as a template because of its high GMQE (0.82) value and identity score (100), with resolution 2.2 Å to build the 3D models of SRC. The PDB structure (ID: 2H8H) of this template has previously been used to identify selective SRC inhibitors.72, 73 Also, the protein structure containing each of the six missense SNPs was individually predicted and visualized using UCSF Chimera 1.17.3 (Fig. 5). Wild-type and mutant models were evaluated using various parameters provided by the validation tools employed. Wild-type has a QMEAN Z-score of −0.24, whereas mutant SRC models ranged from −0.33 to −0.06.

Fig. 5.

Comparative visualization of SWISS-MODEL predicted SRC models through superimposition. (a) 3D structure of the wild-type SRC protein displayed in gold. Superimposed wild-type and mutant structures are shown for the following point mutations, such as (b) W151C, (c) G277D, (d) R388Q, (e) Y419N, (f) P465S, and (g) R483Q with mutants represented in cyan color. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Further validation was performed utilizing SAVES server analysis, which revealed that the wild-type protein had an ERRAT score of 97.8923, a Verify3D score of 92.36 %, and 91.3 % of its residues were placed in the Ramachandran plot’s most favoured regions. All mutant models also showed good structural quality, with more than 90 % of residues in favored areas and high ERRAT (97.1963–98.1265) and Verify3D (91.69 %-92.36 %) scores. As shown in Table 2, additional validation using MolProbity and ProSA provided favorable results for both the wild-type and mutant structures.

Table 2.

Assessment of the quality of SRC wild-type and mutant 3D structures.

Variants	SWISS-MODEL		SAVES (v6.0)			MolProbity score	ProSA Z-score
	Template	QMEAN Z-score	Residues in most favoured regions (percentage)	ERRAT score	Verify3D score
Wild Type	2h8h.1.A	−0.24	358 (91.3 %)	97.8923	92.36 %	1.26	−11.05
W151C	2h8h.1.A	−0.15	360 (91.8 %)	97.6636	92.36 %	1.27	−11.05
G277D	2h8h.1.A	−0.33	361 (91.9 %)	97.6798	92.36 %	1.27	−11.04
R388Q	2h8h.1.A	−0.06	359 (91.6 %)	97.8873	91.69 %	1.26	−11.06
Y419N	2h8h.1.A	−0.06	359 (91.6 %)	97.8923	92.13 %	1.29	−10.97
P465S	2h8h.1.A	−0.15	361 (91.9 %)	98.1265	92.36 %	1.26	−11.04
R483Q	2h8h.1.A	−0.12	359 (91.6 %)	97.1963	92.13 %	1.27	−11.07

3.6. Structural consequences of domain-specific missense nsSNPs

The domains of the SRC protein were analyzed using the InterPro database, which identified three functional domains: Src Homology 3 (SH3; residues 84–145), Src Homology 2 (SH2; residues 147–247), and protein kinase (residues 260–536). Positional analysis of six pathogenic missense SNPs revealed that W151C is located within the SH2 domain, whereas the remaining variants, G277D, R388Q, Y419N, P465S, and R483Q, are situated in the kinase domain. The localization of these variants within the SH2 and kinase domains, regions essential for SRC’s regulatory and catalytic functions, suggests that these mutations have a high potential to disrupt protein activity and promote oncogenic signaling.

Furthermore, the structural impact of six nsSNPs on SRC protein was determined using the HOPE server (Table 3). The analysis revealed significant alterations in residue size, hydrophobicity, and charge, each of which could disrupt protein structure. Mutations that introduce bigger residues than the wild-type, notably G277D is expected to cause steric clashes within the protein core, resulting in structural destabilization. Meanwhile, three nsSNPs, including G277D, Y419N, and P465S, reduced the hydrophobicity of the SRC protein, likely resulting in destabilizing repulsive interactions. Charge-altering mutations such as G277D, R388Q, and R483Q were found to disrupt electrostatic interactions necessary for maintaining the native conformation, compromising SRC protein stability.

Table 3.

Predicted structural alterations in the native SRC protein induced by missense nsSNPs.

Variant	Wild-Type residue			Mutant residue			Prediction
	Residue size	Residue Charge	Hydrophobicity	Residue size	Residue charge	Hydrophobicity
W151C	Bigger	−	−	Smaller	−	−	Damaging
G277D	Smaller	Neutral	More hydrophobic	Bigger	Negative	Less hydrophobic	Damaging
R388Q	Bigger	Positive	−	Smaller	Neutral	−	Damaging
Y419N	Bigger	−	More hydrophobic	Smaller	−	Less hydrophobic	Damaging
P465S	Bigger	−	More hydrophobic	Smaller	−	Less hydrophobic	Damaging
R483Q	Bigger	Positive	−	Smaller	Neutral	−	Damaging

Further analysis using the Missense3D tool revealed damaging effects for three of the twelve analyzed variants. Among these, W151C was particularly disruptive, as it disrupts all hydrogen bonds formed by the wild-type tryptophan, with a relative solvent accessibility of 6.6 %. Additionally, the Y419N and P465S mutations were predicted to expand surface pocket volumes by 88.128 ų and 75.6 ų, respectively, potentially altering ligand binding or protein interactions. Energy minimization using SwissPdbViewer provided total energy scores of −25125.572 kJ/mol for the wild-type structure, compared to −24959.918 kJ/mol (W151C), −25180.844 kJ/mol (Y419N), and −25189.281 kJ/mol (P465S). These structures were then subjected to MD simulations and docking analyses.

3.7. Molecular dynamics simulations analysis

Proteins are dynamic molecules that constantly undergo conformational changes in a cellular environment. These movements are not random but are intrinsically tied to their biological functions. MD simulations for the three most detrimental SNPs (W151C, Y419N, and P465S) were conducted for 200 ns to analyze the structural alterations of SRC protein caused by these nsSNPs under normal physiological conditions.

3.7.1. Stability analysis

RMSD values of the carbon alpha (C $\mathrm{\alpha }$) atoms were analyzed to evaluate the structural deviations of the SRC protein and its mutants over time during the simulation. It provides insights into protein stability, wherein smaller RMSD values imply enhanced stability.⁷⁴ RMSD trajectories for the wild-type SRC protein and its mutant variants (W151C, Y419N, and P465S) were plotted against the simulation time (Fig. 6A). The average RMSD values observed were 2.93 Å for W151C, 3.42 Å for Y419N, and 3.07 Å for P465S, while the wild-type protein exhibited a lower average RMSD of 2.68 Å. The maximum RMSD deviations recorded were 4.03 Å for the wild-type, 4.07 Å for W151C, 4.54 Å for Y419N, and 4.12 Å for P465S, whereas minimum RMSD value was observed for wild-type (1.35 Å), followed by W151C (1.37 Å), Y419N (1.39 Å), and P465S (1.82 Å). The results indicated that all mutants exhibited higher RMSD values and fluctuations during simulation, suggesting less stability than native protein.

Fig. 6.

200 ns simulation trajectories of the wild-type SRC and three mutant variants. (A) SRC C $\mathrm{\alpha }$ RMSD analysis. (B) RMSF properties. (C) Rg analysis. (D) Intramolecular hydrogen bonding analysis.

3.7.2. Flexibility analysis

RMSF is an essential metric in MD simulations and is used to assess the stability and flexibility of complex systems; higher RMSF values denote increased flexibility.⁷⁵ The structural flexibility of the wild-type protein and its three mutants was evaluated by calculating RMSF values. Distinct residues exhibited varying peak changes in each protein (Fig. 6B). For the wild-type protein, notable fluctuations were detected at residues 438–441. In the W151C mutant, variations were identified at residues ALA_197 and SER_212 and within the range of residues 524–527. The P465S mutant exhibited substantial changes in residues THR_182, GLY_291, THR_292, GLY_424, ALA_425, and GLY_462. The Y419N mutant exhibited fluctuations in GLN_423, GLY_424, TYR_466, PRO_467, VAL_470, ASN_471, and ARG_472. The mean RMSF values for the wild type, W151C, Y419N, and P465S proteins were 1.282, 1.456, 1.6, and 1.35 Å, respectively. These results indicated that all mutant proteins exhibited increased flexibility compared to the wild-type protein, accompanied by significant residual changes.

3.7.3. Compactness analysis

Radius of gyration (Rg) is a crucial measure for analyzing the compactness of a protein, which quantifies the root-mean-square distance between the protein's atoms and its axis of rotation. Proteins with low Rg values are more compact and rigid, whereas those with high values are less compact and flexible.75, 76 To evaluate the structural integrity of the wild-type SRC protein and its mutant variants, the time evolution of Rg was analyzed (Fig. 6C). Wild-type SRC protein exhibited an average Rg value of 25.431 Å, indicating a stable and well-packed structure throughout the simulation. In contrast, W151C, Y419N, and P465S showed slightly lower average Rg values of 24.345, 24.387, and 24.281 Å, respectively. The results showed that all mutants had more compact and less flexible structures, since their Rg values were lower than the native structure.

3.7.4. Hydrogen bond analysis

To comprehend the structural impact of the three most deleterious missense SNPs in the SRC protein, we conducted a comparative analysis of their influence on hydrogen bonding patterns. As shown in Fig. 6D, the number of hydrogen bonds fluctuated over time in both the wild-type and mutant forms of SRC during the simulations. W151C, Y419N, and P465S variants exhibited average hydrogen bonds of 383, 395, and 385, respectively, indicating marginal changes from the wild-type average of 383. These changes in hydrogen bonds can affect the structure of the SRC protein and lead to abnormal functioning.

3.7.5. Secondary structure element analysis

To identify the alteration in SRC protein secondary structural arrangement caused by the three most damaging nsSNPs, the secondary structure elements ($\mathrm{\alpha }$-helix and $\mathrm{\beta }$-sheets) of wild-type and mutant proteins were analyzed, as shown in Supplementary Fig. S2. Native SRC protein has a total of 39.01 % secondary structure composition, with 20.33 % $\mathrm{\alpha }$-helix and 18.68 % $\mathrm{\beta }$-sheets. In contrast, W151C mutant has a total secondary structure of 37.35 %, comprising 18.41 % α-helix and 18.94 % $\mathrm{\beta }$-sheets. Meanwhile, Y419N mutant shows an increased total of 40.73 %, featuring 20.57 % $\mathrm{\alpha }$-helix and 20.17 % $\mathrm{\beta }$-sheets, whereas P465S mutant also surpasses the native protein, exhibiting a total secondary structure of 40.10 %, which includes 20.28 % $\mathrm{\alpha }$-helix and 19.82 % $\mathrm{\beta }$-sheets. The analysis shows that the SSE of W151C decreases in overall composition, while the Y419N and P465S mutants show a higher total SSE than the original SRC protein. These variations in SSE indicate the influence of nsSNPs on SRC protein structure.

3.7.6. Principal component analysis (PCA) and dynamics cross-correlation matrices (DCCM) analysis

PCA evaluated the effects of three harmful missense SNPs on the structural dynamics and fluctuations of both wild-type and mutant proteins. Analyzing trajectory motions during MDS provides insights into protein behavior. The eigenvalues in these simulations indicate fluctuations along hyperspace eigenvectors, where eigenvectors with higher eigenvalues play a crucial role in regulating overall protein mobility.⁷⁷ In the wild-type protein, the top five eigenvectors displayed dominant motions, accounting for 23.4 % to 66.5 % of the total fluctuations. Likewise, in the W151C, Y419N, and P465S mutants, the top five eigenvectors demonstrated dominant movements, with eigenvalues ranging from 26.1 % to 70.4 %, 44.2 % to 78.8 %, and 34.9 % to 67.3 %, respectively. More than 50 % of the structural variance was captured by the first three principal components (PC1, PC2, and PC3). For the wild-type, PC1 accounted for 23.39 % of the variance, followed by PC2 at 11.96 % and PC3 at 11.42 %, summing to 46.77 % (Fig. 7A). In individual cases, W151C showed variance contributions of 26.05 % for PC1, 16.75 % for PC2, and 11.44 % for PC3 (Fig. 7B). Y419N presented 44.19 % for PC1, 15.59 % for PC2, and 5.77 % for PC3 (Fig. 7C). P465S revealed 34.87 % for PC1, 11.35 % for PC2, and 6.83 % for PC3 (Fig. 7D). These indicate that while PC1 shows the largest portion of variance across all cases, the relative contributions of PC2 and PC3 differ, reflecting structural alteration between variants.

Fig. 7.

Principal Component Analysis (PCA, left) and Dynamic Cross-Correlation Matrix (DCCM, right) of wild-type SRC (A) and mutant forms W151C (B), Y419N (C), and P465S (D). The PCA plots depict the distribution of eigenvalues relative to the percentage of variance, indicating the distribution of conformational motion. Corresponding DCCM plots illustrate residue-residue dynamic correlations, with a color gradient ranging from light sea green (negative correlation) to dark blue (positive correlation). (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

The arrangement of blue and red dots illustrates conformational shifts during the simulation, with the intensity of color reflecting structural mobility. Blue indicates the highest level of mobility, while white and red denote moderate and minimal movement, respectively. These patterns demonstrate apparent clustering in the principal component (PC) subspace, emphasizing the conformational differences across all groups. Significantly, the mutations impact the native conformation, as none of the mutants exhibit movement patterns akin to those of the wild-type.

Additionally, DCCM analysis was done for native SRC and its mutants (W151C, Y419N, P465S) to investigate correlated and anti-correlated residues. Fig. 7A-D shows dark blue areas for highly correlated residues and sea green for weakly correlated ones. Each plot has a dark blue diagonal line, indicating the self-correlation of residues. The wild-type SRC shows a balanced correlation pattern with moderately correlated regions throughout. The W151C mutation reduces correlation intensity, seen in the expanded sea green areas, indicating weaker interactions. Y419N disrupts domain interactions, creating a distinct separation between matrix quadrants. P465S shows the highest deviation, with uniform cyan regions and loss of finer structural details, suggesting widespread moderate correlations and structural disruption.

3.8. SRC gene-gene interaction network analysis

GeneMANIA and STRING database analyses identified genes and proteins interacting with SRC within a complex biological network. Significant genes that interact with SRC, according to GeneMANIA (Fig. 8A), include sorbin and SH3 domain-containing 1 (SORBS1), subunits of RNA polymerase, protein tyrosine kinase 2 (PTK2), epidermal growth factor receptor (EGFR), C-Src kinase (CSK), HRas proto-oncogene, GTPase (HRAS), Erb-b2 receptor tyrosine kinase 2 (ERBB2), receptor for activated C kinase 1 (RAK1), and cell division cycle 42 (CDC42). In line with this, STRING (vs. 12) identified ten major proteins interacting with the SRC protein. These include platelet-derived growth factor receptor beta (PDGFRB), E3 ubiquitin-protein ligase CBL (CBL), signal transducer and activator of transcription 3 (STAT3), paxillin (PXN), receptor tyrosine-protein kinase erbB-2 (ERBB2), epidermal growth factor receptor (EGFR), adapter molecule Crk (CRK), talin-1 (TLN1), hematopoietic lineage cell-specific protein (HCLS1), and heat shock protein HSP 90-alpha (HSP90AA1) (Fig. 8B). A list of major partners associated with the SRC interaction network, as detailed in the STRING (version 12) database, is provided in Supplementary Table S3. The interacting biomolecules identified by these tools suggest that SRC plays a role in several pathways. Polymorphisms in the SRC gene can result in changes to protein–protein and gene-gene interaction networks. These genetic variations may disrupt the normal functioning of the SRC pathway, leading to significant implications for the molecular mechanisms that underlie various diseases.

Fig. 8.

The interaction network of SRC with other genes and proteins was analyzed using A) GeneMANIA and B) STRING.

3.9. Molecular docking analysis

Molecular docking analyses demonstrated distinct alterations in binding affinities between the wild-type SRC and its variants (W151C, Y419N, and P465S) when interacting with FAK and the kinase inhibitor dasatinib. ClusPro 2.0 was employed to investigate the SRC and FAK interactions. Wild-type SRC-FAK complex exhibited binding affinity, with a docking score of −849.1. W151C-FAK and Y419N-FAK have lower binding affinities than wild-type, scoring −757.5 and −718.8, respectively. In contrast, the P465S-FAK complex had a higher binding affinity (−850.6) than the wild type. Furthermore, PDBsum revealed that wild-type-FAK featured 7 salt bridges, 18 hydrogen bonds, and 220 non-bonded contacts, while missense SNPs led to changes in these interactions, as shown in Fig. 9. Salt bridges are specific interactions involving charged atoms that facilitate binding; hydrogen bonds produce precise links between molecules that resemble small bridges; and non-bonded contacts increase the overall attraction. mCSM-PPI2 predictions provided insight into the stability changes induced by each mutation within the SRC-FAK complex, where W151C-FAK showed the most substantial destabilization (−0.625 kcal/mol), followed by Y419N (−0.593 kcal/mol) and P465S (−0.198 kcal/mol), as summarized in Table 4. A ΔΔG (kcal/mol) < 0 indicates a lower binding affinity. Y419N significantly impacts SRC-FAK binding, reducing affinity and destabilizing the complex, while P465S causes a slight change in binding interactions.

Fig. 9.

Depiction of the SRC (chain A) and FAK (chain B) complex interaction with its mutants. (A) Wild-type SRC-FAK, (B) W151C-FAK, (C) Y419N-FAK, and (D) P465S-FAK, highlighting alterations in interacting residues. FAK is shown in blue, while wild-type, W151C, Y419N, and P465S are depicted in red, green, cyan, and purple, respectively. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Table 4.

The impacts of mutations on SRC–FAK interactions and binding affinity.

Docking Complex	ClusPro 2.0	PDBsum			mCSM-PPI2
	Lowest Energy	No. of Salt bridges	No. of Hydrogen bonds	No. of non-bonded contacts	ΔΔG (kcal/mol)	Affinity change
Wild Type-FAK	−849.1	7	18	220	−	−
W151C-FAK	−757.5	5	23	183	−0.625	Decreasing
Y419N-FAK	−718.8	−	4	249	−0.593	Decreasing
P465S-FAK	−850.6	7	18	221	−0.198	Decreasing

Further docking analysis using PyRx was performed to determine the binding patterns of dasatinib with wild-type and mutant SRC proteins. To ensure the reliability of the docking protocol, the co-crystallized ligand quinazoline was extracted from the crystal structure using Discovery Studio and redocked into the SRC binding site. The redocked ligand consistently occupied the same binding pocket as the native ligand, replicating interactions with residues LEU273, VAL281, ALA293, LYS295, GLU339, TYR340, and LEU393, which are similar to the original co-crystal structure. Following protocol validation, dasatinib was docked with the SWISS-MODEL-generated SRC wild-type and mutant models (Supplementary Fig. S3). Docking scores, RMSD values, and grid box parameters are detailed in Supplementary Table S4.

In wild-type SRC, dasatinib demonstrated the highest binding affinity (−9.2 kcal/mol), stabilized by six hydrogen bonds with LYS298 (2), PHE408 (2), and ASP407 (2), as well as nine hydrophobic interactions involving VAL284 (2), LEU396, ALA296 (2), LYS298, ALA406, LEU276, and TYR343. For the mutants, W151C-dasatinib showed a slightly lower affinity (−9.0 kcal/mol), forming three hydrogen bonds at ARG391 (3) and ASP407, two electrostatic interactions at LYS298, ARG391, and ten hydrophobic interactions at VAL284 (3), LEU396 (3), ALA296 (2), LYS298, and ALA406 (2). Y419N-dasatinib also had a docking score of −9.1 kcal/mol, generating three hydrogen bonds with ARG391, ASP407 (2), two electrostatic interactions with LYS298 and ASP389, along with eight hydrophobic contacts with LEU276, PHE281, VAL284 (2), LEU396, ALA406, ALA411, and ILE414. P465S-dasatinib had a docking score of −9.0 kcal/mol, with one hydrogen bond at ASP407, one electrostatic bond at ARG391, and eleven hydrophobic interactions at VAL284 (3), LEU396 (2), ALA296 (2), LYS298, LEU276, and ALA406 (2). The interacting residues of SRC-dasatinib complexes are shown in Fig. 10. All SRC mutants showed decreased binding affinity and changed interacting residues compared to the wild-type.

Fig. 10.

Visualization of the altered interacting residues of SRC-dasatinib complexes. Panels (A-D) illustrate interactions with Dasatinib, (A) Wild-type, (B) W151C, (C) Y419N, and (D) P465S.

3.10. Determining the cancer-associated high-risk nsSNPs

Twelve harmful missense nsSNPs were evaluated using CScape, Dr. Cancer, and CancerVar. According to CScape, all twelve missense nsSNPs have oncogenic potential, indicating a possible association with cancer. CancerVar identified ten nsSNPs as oncogenic, while Dr. Cancer identified seven SNPs as disease-associated and five as neutral. The agreement among the three analytical tools regarding seven of the twelve SNPs enhances the reliability of their oncogenic implications (Table 5). When taken as a whole, these variations highlight how crucial they are in regulating cancer risk. Importantly, all three tools consistently identified W151C, Y419N, and P465S mutations as oncogenic.

Table 5.

Shortlisted missense nsSNPs with oncogenic potential.

Mutation	CScape		Dr. Cancer		CancerVar
	Score	Prediction	Score	Prediction	Score	Prediction
W151C	0.915895	Oncogenic (high confidence)	0.532	Disease	0.96	Oncogenic
K203E	0.583531	Oncogenic	0.757	Disease	0.88	Oncogenic
G277D	0.856495	Oncogenic	0.482	Neutral	0.92	Oncogenic
A374T	0.887398	Oncogenic	0.817	Disease	0.89	Oncogenic
V386D	0.901357	Oncogenic (high confidence)	0.716	Disease	0.92	Oncogenic
R388Q	0.722843	Oncogenic	0.621	Disease	0.92	Oncogenic
Y419N	0.748828	Oncogenic	0.704	Disease	0.87	Oncogenic
G440D	0.884838	Oncogenic	0.372	Neutral	0.78	Oncogenic
I444N	0.899857	Oncogenic (high confidence)	0.365	Neutral	0.51	Unknown
P465S	0.892162	Oncogenic (high confidence)	0.637	Disease	0.91	Oncogenic
R483Q	0.923052	Oncogenic (high confidence)	0.329	Neutral	0.87	Oncogenic
M484V	0.850368	Oncogenic	0.340	Neutral	0.39	Unknown

3.11. Regulatory effects of SRC and associated gene mutations in cancer

SRC expression was analyzed using muTarget, a tool that identifies genes whose expression is significantly affected by mutated genes (and vice versa) across different cancer types. It can associate genes and their mutational influence on gene expression for more than 30 genes, of which the top ten were chosen. The ‘Genotype’ analysis revealed that SRC mutations significantly affect the expression of various genes involved in multiple myeloma and uterine cancer. As shown in Supplementary Table S5, Fig. S4, SRC mutations were linked to changes in the expression of genes such as ADM2, ARHGAP17, HSH2D, LPGAT1, and ZFPM1 in multiple myeloma, as well as FAM3B, HSD3B7, KLHL21, MAPK12, and MINK1 in uterine cancer.

The ‘Target’ analysis identified mutant genes influencing SRC expression across various cancer types, including lung, liver, bladder, breast, multiple myeloma, uterine cancer, and colon adenocarcinoma. The identified gene mutations include CBFB in breast cancer; KMT2D, SPEG, BAP1, PTPN14, and CSMD2 in liver cancer; TRIP12, PRKDC, CHD9, AKAP12, and PRDM2 in colon adenocarcinoma; RB1, FGFR3, ERBB2IP, BTBD7, and FLNA in bladder cancer; GUCY2F, IGKV2-30, IRF4, SLC22A3, and TG in multiple myeloma; WIPF3, MOAP1, IL10, SLC35A4, and NIPAL4 in uterine cancer; TPK1, GRIK1, DDI1, ZNF780B, and CD200R1 in lung adenocarcinoma (Supplementary Table S6, Fig. S5).

4. Discussion

SRC is a non-receptor tyrosine kinase that regulates key oncogenic processes, including uncontrolled cell proliferation, survival, migration, and angiogenesis.⁷⁸ By integrating signals from receptor tyrosine kinases and other ligand-mediated pathways, SRC activates downstream cascades such as PI3K-AKT, Ras-MAPK, JAK-STAT3, and FAK-paxillin, which collectively driving cellular invasion, metastasis, and drug resistance.79, 80 Although much research has focused on dysregulated SRC signaling, the structural and functional consequences of SRC mutations remain poorly characterized. In this study, we investigated potentially pathogenic missense SNPs in the SRC gene using an integrative in silico pipeline that combined variant prediction, three-dimensional structural modeling, molecular dynamics simulations, molecular docking, and expression analysis. This cumulative strategy, which incorporates multiple algorithms and scoring systems, reduces methodological bias and enhances prediction accuracy, thereby enabling a more comprehensive assessment of SRC variants.

Our analysis began by retrieving 512 missense SNPs in the SRC gene from the NCBI dbSNP database. To identify pathogenic and destabilizing variants, we employed a consensus-based approach, integrating multiple prediction tools and cross-validating their results. Following this screening, 42 SNPs were predicted as deleterious by nine prediction tools. These variants were subsequently evaluated using seven structure- and sequence-based tools to determine their effects on protein stability. Protein stability is important for functional integrity because it determines whether a protein remains properly folded or unfolds into a nonfunctional, denatured state.⁸¹ Out of the 42 pathogenic missense SNPs, 12 variants (W151C, K203E, G277D, A374T, V386D, R388Q, Y419N, G440D, I444N, P465S, R483Q, and M484V) were predicted to decrease protein stability. Altered protein stability can result in misfolding, disintegration, or abnormal aggregation, ultimately impairing protein function and biological processes.⁸² Notably, six of these mutations occurred at highly conserved residues, including positions shared with closely related primates, highlighting their evolutionary importance. Because mutations in evolutionarily conserved residues are often pathogenic and can disrupt normal physiological functions, these destabilizing SNPs may play a role in cancer development.⁸³

Subsequently, our analysis proceeded by predicting the 3D structure of the SRC wild-type and mutants using SWISS-MODEL. Both native and mutant models were constructed from SWISS-MODEL using the “2h8h.1.A” template for structural analysis. To further explore the structural consequences of these variants, predictive analyses were carried out using HOPE and Missense3D. HOPE analysis revealed changes in size, hydrophobicity, and charge across six missense SNPs. Also, all variants were detected in the SH2 regulatory and protein kinase catalytic domains, both essential for SRC function. Missense3D analysis predicts the three most pathogenic variants (W151C, Y419N, and P465S) influencing protein structure out of the six SNPs. SRC has distinct domains that are essential for its regulatory and enzymatic actions. The SH3 and SH2 domains facilitate protein–protein interactions, allowing SRC to bind to signaling molecules, whilst the kinase domain phosphorylates tyrosine residues on target proteins, altering their activities.⁸⁴ The W151C mutation occurs within the SH2 domain, indicating the possibility of signal transduction pathway abnormalities. Conversely, mutations at Y419N and P465S are located in the protein-tyrosine kinase domain, which is critical for catalytic activity. Phosphorylation at Tyr419 is a regulatory mechanism that determines the functional state of SRC.⁸⁵ Mutations at Tyr419, such as Y419N, can interfere with SRC function and contribute to aberrant expression levels.

MD simulations offer valuable insights into critical processes, such as protein folding, conformational changes, ligand interactions, and mutation effects, providing a dynamic perspective that static models cannot capture.⁸⁶ To further analyze the wild type and W151C, Y419N, and P465S mutations, a 200 ns MD simulation study was conducted. RMSD analysis demonstrated that the mutants decrease protein stability, which perfectly aligned with the result obtained from static stability prediction tools. RMSF and Rg analyses showed variations in structural flexibility and compactness of the mutant proteins compared to the native. Protein folding is crucial for maintaining functional integrity because it makes it easier for $\mathrm{\alpha }$-helices and $\mathrm{\beta }$-sheets to form, which promote higher-order structures.⁸⁷ SRC 2D structure analysis uncovered significant differences among the mutants during simulation. P465S and Y419N mutations increased $\mathrm{\alpha }$-helices and $\mathrm{\beta }$-strands, whereas the W151C mutation decreased these structural elements, indicating a marked alteration in the protein’s structural configuration. Furthermore, principal component analysis revealed that the wild-type and mutant proteins clustered differently, suggesting that the mutations may impair the protein's ability to adopt its native conformation. Meanwhile, DCCM analysis revealed disruptions in long-range residue interactions, which could affect enzyme activity and allosteric regulations, resulting in functional deficiencies.

Molecular docking analyses were performed to evaluate the impact of nsSNPs on SRC protein and FAK interaction, as well as its inhibitor, the anticancer drug dasatinib. SRC, a key regulator of cancer metastasis, modulates essential cellular processes such as cytoskeletal organization, cell proliferation, and survival by phosphorylating FAK.⁸⁸ Mutations in the SRC can disrupt these interactions, resulting in abnormal signaling that promotes tumor progression. Particularly in bladder cancer, TGF-β-induced cell migration and invasion depend on the SRC-mediated regulation of FAK.⁸⁹ Docking of SRC and FAK using ClusPro2.0, which ranks models based on cluster size and energy scores, revealed that the Y419N mutation reduced binding affinity, with the lowest energy score decreasing from −849.1 (wild type) to −718.8, indicating a destabilized SRC-FAK interaction. PDBsum analysis showed that Y419N disrupts hydrogen bonding and salt bridges at the interface. W151C also reduced binding affinity (−757.5), while P465S slightly increased it (−850.6); nonetheless, mCSM-PPI2 predicted decreased interaction for all variants, underscoring their potential to impair SRC-FAK complex stability.

In the protein–ligand docking investigation, wild-type SRC had a docking score of −9.3 kcal/mol with dasatinib, while the Y419N mutant had a slightly lower score of −9.2 kcal/mol. The W151C and P465S variants both showed a decrease in binding affinity, with scores of −9.1 kcal/mol. Although the energy differences were marginal, Y419N is particularly noteworthy because Tyr419 is the critical autophosphorylation site required for SRC activation.⁹⁰ In addition to the changes in binding energy, all three mutations altered hydrogen bonding patterns and hydrophobic interactions at the ligand-binding site. These findings indicate that nsSNPs, particularly W151C, Y419N, and P465S, may impair SRC-FAK and SRC-dasatinib interactions, potentially leading to increased disease susceptibility and therapeutic resistance.

Further assessment of nsNPs has identified seven variants in the SRC gene, including W151C, Y419N, and P465S, as oncogenic. These mutations may contribute to cancer pathogenesis, given that SRC is a well-known proto-oncogene. In addition to this, muTarget analysis revealed several genes regulating SRC expression, as well as genes influenced by SRC mutations, in multiple myeloma and uterine cancer. Of them, mutated CBFB is one of the critical genes linked to the development of breast cancer.⁹¹ These findings demonstrate the complexity of SRC regulation in cancer and highlight the potential of associated genes as therapeutic targets. Elucidating this regulatory network may provide new avenues for intervention across diverse malignancies.

This study used a comprehensive in silico approach to evaluate pathogenic missense nsSNPs in the SRC gene, integrating functional, structural, and genomic data. However, some drawbacks should be acknowledged. The analysis, which is solely computer-based, cannot adequately represent the complexity of biological systems. Molecular dynamics simulations and docking analyses rely on approximate force fields and energy functions, which cannot fully capture the complexities of SRC-FAK or SRC-dasatinib interactions in the cellular environment. Additionally, the predictive accuracy of our findings is influenced by the quality and limitations of the underlying bioinformatics tools and their training datasets. While we observed possible associations between certain nsSNPs and cancer susceptibility, their prevalence in the broader population and clinical significance remain unclear. Further epidemiological investigations are needed to determine their frequency among cancer patients versus healthy controls in order to explore their potential as biomarkers for early detection, risk stratification, or therapeutic intervention.

Despite these limitations, our findings highlight the structural and functional impact of SRC missense variants, particularly W151C, Y419N, and P465S. By destabilizing protein structure, disrupting signaling interactions, and potentially altering drug binding, these variants may promote tumorigenesis and therapy resistance. Therefore, follow-up studies should validate these predictions in cellular systems and preclinical models to evaluate their oncogenic potential and translational relevance in precision oncology. Functional assays, along with targeted in vitro and in vivo testing, will be essential to confirm the pathogenicity and therapeutic implications of these variants.

5. Conclusion

This study presents a computational exploration of SRC missense mutations, highlighting their deleterious impact on protein stability, conformational dynamics, and molecular interactions. Through an integrative approach combining structural modeling, molecular dynamics simulations, and docking, we show that the pathogenic variants W151C, Y419N, and P465S induce significant structural perturbations that compromise SRC’s functional integrity. These mutations impair regulatory interactions with FAK and alter the binding affinity of the clinically relevant kinase inhibitor Dasatinib, underscoring mutation-specific differences in therapeutic responsiveness. These findings lay the groundwork for utilizing SRC mutational status in precision medicine, aiding in patient stratification and therapy choices. For effective clinical application, future research should prioritize population-scale screening of SRC variants, assess their functional consequences in diverse patient cohorts, and validate their roles in cancer progression and drug resistance.

Data Availability Statement

All data will be available upon request to the corresponding author.

Funding Source

None. No fund was received for this study.

CRediT authorship contribution statement

Md. Shakil Ahamed: Writing – original draft, Validation, Methodology, Formal analysis, Data curation, Conceptualization. Roksana Khanam: Writing – review & editing, Validation, Methodology, Formal analysis, Data curation. K.M. Tanjida Islam: Validation, Methodology, Formal analysis, Data curation. Fahmida Tabassum: Validation, Methodology, Formal analysis, Data curation. Md. Al Amin: Validation, Methodology, Formal analysis, Data curation. Jannatul Fardous: Validation, Methodology, Formal analysis, Data curation. Nadira Hoque Tashpie: Validation, Methodology, Formal analysis, Data curation. A.K.M. Mohiuddin: Writing – review & editing, Validation, Methodology, Formal analysis. Shahin Mahmud: Writing – review & editing, Supervision, Methodology, Investigation, Formal analysis, Conceptualization.

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: