The present study focused on five germline variants in cancer-predisposing genes that have been classified as VUS. Comprehensive bioinformatics analysis using multiple in-silico computational tools provided a better understanding of the potential pathogenicity, stability, functional impact, protein–protein interactions, and structural alterations caused by these variants. MDS provided further insights into the dynamic behavior of the wild-type and variant proteins, highlighting potential effects on protein stability and flexibility.

Heterogeneous results were obtained regarding the potential pathogenicity of the five VUS. This highlights the inherent limitations of these tools, as evidenced by conflicting predictions for certain variants. For example, the BRCA1:(NM_007294.4):c.3392A > G;p.Asp1131Gly variant displayed discrepancies between different pathogenicity prediction tools. The ClinVar database indicates five submissions for this variant, including four clinical testing submissions (uncertain significance) and one curation submission (likely benign) [36]. While one submission suggested a non-conservative amino acid change and three out of five in-silico tools predicted a damaging effect, the available data on variant occurrences in the general population are insufficient to draw definitive conclusions about its significance. The variant has been reported in at least one individual affected with Hereditary Breast and Ovarian Cancer Syndrome, but strong evidence for causality is lacking [37,38,39]. Dines et al. suggested that this variant might be located in a "coldspot" region, where missense variants are less likely to be pathogenic. However, other studies have indicated that the available evidence is insufficient to determine the role of this variant in disease. Additionally, algorithms developed to predict the effect of sequence changes on RNA splicing suggest that this variant may create or strengthen a splice site [40]. Advanced modeling of protein sequence and biophysical properties performed at Invitae suggests that this missense variant is not expected to disrupt BRCA1 protein function [36]. The amino acid change at codon 1131 replaces aspartic acid with glycine, which have similar properties, and the amino acid position is not well conserved in available vertebrate species. Furthermore, in-silico predictions for this alteration are inconclusive [35]. Based on the available evidence, including the conflicting in-silico predictions, limited population data, and lack of strong experimental evidence, the clinical significance of the BRCA1:(NM_007294.4):c.3392A > G;p.Asp1131Gly variant remains uncertain. Further research, including functional genomic studies, is warranted to definitively assess its role in breast and ovarian cancer susceptibility.

Similarly, the BRIP1:(NM_032043.3):c.3103C > T;p.Arg1035Cys, variant yielded conflicting results in the current study, with some tools classifying it as non-pathogenic and others suggesting deleterious potential (PredictSNP: 51% score, Align GVGD: C65 class with a high chance of deleterious effects). These findings underscore the crucial need for a multifaceted approach that combines in-silico analysis with additional methods for more definitive assessment of the functional relevance. The ClinVar database has 19 submissions for the BRIP1:(NM_032043.3):c.3103C > T;p.Arg1035Cys variant. Eighteen of them were clinical testing and one was a curation submission. Nine of those submissions reported the BRIP1 variant is of uncertain significance, while six reported it as likely benign. Four submissions had reported it as a benign variant based on a combination of the following: variant is present in unaffected individuals, population frequency, intact protein function, lack of segregation with disease, co-occurrence, RNA analysis, in-silico models, amino acid conservation, lack of disease association in case–control studies, and/or the mechanism of disease or impacted region is inconsistent with a known cause of pathogenicity [41].

The MET:(NM_000245.4):c.840G > T;p.Arg280Ser variant was predicted to have high-risk potential for altering protein structure and function. These include a possible disruption of disordered regions (altered disordered interface, 0.31 probability), changes in metal binding (altered metal binding, 0.28 probability), and a slight possibility of affecting its role as a transmembrane protein (altered transmembrane protein, 0.13 probability). Interestingly, the analysis also suggested a potential gain of a catalytic site at residue R277 (0.13 probability). Additionally, some alterations in protein motifs were observed. These findings warrant further investigation to determine the precise functional consequences of the variant on the MET protein. The ClinVar database has four clinical testing submissions for the MET:(NM_000245.4):c.840G > T;p.Arg280Ser variant and each of them reported the variant as a VUS [42]. It is noteworthy that while the MET gene has not previously been linked to hereditary breast cancer in the literature, this variant was identified in our exome database from a 42-year-old female diagnosed with high-grade ductal breast carcinoma in situ. This patient's tumor was triple-positive (ER + , PR + , and HER2 +). Additionally, the patient had a family history of breast cancer, with affected first-, second-, and third-degree relatives. It was based on this compelling clinical context that the variant was selected for further investigation in this study.

The CHEK2:(NM_007194.4):c.60G > T;p.Gln20His variant suggested an intermediate impact. All three ClinVar submissions for this variant have identified it as a VUS [43]. The STK11:(NM_000455.5):c.355A > G;p.Asn119Asp suggested no significant structural or functional impact on the protein. The ClinVar record has eight submissions for the STK11:(NM_000455.5):c.355A > G;p.Asn119Asp variant. Seven of them were clinical testing and one was a curation submission. Seven of those submissions reported the STK11 variant is of uncertain significance, while one reported it as likely benign. According to the Ambry Genetics entry, “The asparagine at codon 119 is replaced by aspartic acid, an amino acid with highly similar properties. This amino acid position is highly conserved in available vertebrate species. In addition, this alteration is predicted to be tolerated by in-silico analysis. Since supporting evidence is limited at this time, the clinical significance of this alteration remains unclear” [44].

3D protein modeling with structural analysis revealed a spectrum of changes created by the five variants. These changes ranged from subtle alterations in the backbone conformation, observed in the BRIP1 protein, to significant modifications in surface features, as observed with the MET protein. This finding suggests potentially varying functional consequences for the different proteins encoded by these variants.

Protein–protein interactions for the BRCA1 and MET proteins were explored using STRING to understand their roles in cellular processes. Studies have shown that BRCA1 interacts with proteins such as PALB2, RBBP8 and BRCA2, which are crucial for facilitating DNA repair through homologous recombination (HR) and inhibiting the error-prone non-homologous end joining (NHEJ) pathway [45]. Additionally, BRCA1 interacts with BARD1, forming the BRCA1/BARD1 complex which acts as an E3 ubiquitin ligase, attaching ubiquitin molecules to specific target proteins. Ubiquitination by BRCA1/BARD1 influences DNA repair, cell cycle control, and gene regulation. Mutations in BRCA1 and BARD1 disrupt the ubiquitin ligase activity of the complex [46].

Similarly, analysis using STRING and Genemania identified HGF as the strongest interactor with the MET protein. This finding is supported by the literature demonstrating that HGF binding activates c-MET, a receptor tyrosine kinase, promoting processes involved in cancer development, such as cell proliferation, migration, and metastasis [47,48,49,50,51]. Interestingly, a study by Papa et al. [52] suggested a potential negative interaction between the TGF-β and HGF pathways, highlighting the complex interplay between signaling pathways in cancer biology. The analysis of protein–protein interactions provides valuable context for interpreting the functional impact of the observed structural alterations in the BRCA1 and MET proteins.

While these in-silico methods provide a valuable starting point, their use is not devoid of limitations. For example, the short simulation timescales employed during MDS might not capture the full range of protein dynamics. Additionally, techniques such as analyzing hydrogen bond dynamics or solvent accessibility were not explored, potentially offering a more nuanced picture.

The structural models for the variant proteins of BRCA1, BRIP1, CHEK2, MET and STK11 as well as the wild-type models of BRCA1, BRIP1, CHEK2 proteins are novel as they were generated de novo through this study. Through a rigorous bioinformatics analysis, out of the five VUS that underwent pathogenicity assessment, we prioritize and recommend three VUS for further experimental validation: MET:(NM_000245.4):c.840G > T;p.Arg280Ser, BRCA1:(NM_007294.4):c.3392A > G;p.Asp1131Gly, and BRIP1:(NM_032043.3):c.3103C > T;p.Arg1035Cys. These variants present with a high-risk potential for affecting protein stability and function, warranting further investigation to delineate their exact biological effects and for a better understanding of their role in breast cancer development. Currently, insufficient evidence precludes the reclassification of these VUS on the basis solely of in-silico analysis. However, further experimental studies demonstrating the impact of the variant on protein function could lead to its reclassification as likely pathogenic or benign. This would provide a more definitive risk assessment for individuals harboring such variants.

It is important to note that in line with the methodology used in our study, several researchers have emphasized the importance of using multiple bioinformatic tools for the pathogenicity prediction of SNV with greater accuracy rather than relying solely on a single tool [3, 4]. Galehdari et al. [53], reporting on the diagnostic accuracy of SNV-based pathogenicity detection tools for UGT1A1 gene variants underscored this point in their meta-analysis. They compared the results of various SNV-based prediction tools with published clinical results for pathogenicity prediction of nonsynonymous SNV associated with Crigler-Najjar syndrome. While some tools like SIFT and PolyPhen-2 demonstrated promising results, the study highlighted the limitations of individual tools in accurately predicting the pathogenicity of variants [53]. Additionally, factors beyond structural stability, such as disruptions in post-translational modifications or ligand binding, can influence the disease phenotype.

Several limitations were identified during this study. The study relied heavily on computational bioinformatics tools, which are subject to limitations and potential biases. Conflicting predictions were observed for certain variants, such as BRCA1:(NM_007294.4):c.3392A > G and BRIP1:(NM_032043.3):c.3103C > T. Furthermore, the absence of experimental validation, both in vitro and in vivo, hinders the confirmation of variant pathogenicity and structural impact, limiting the clinical applicability of these results. Due to the deceased status or unavailability of family members, a formal segregation analysis could not be performed, which would have provided stronger evidence for variant pathogenicity.

Additional limitations include the relatively short 50-ns timeframe of MDS, which may not fully capture long-term protein dynamics or stability changes. The study did not explore hydrogen bond dynamics or solvent accessibility, which could provide a more nuanced understanding of protein-variant interactions. Computational resource constraints precluded the inclusion of MDS in the analysis of the MET variant, resulting in a partial assessment of its stability.

To establish definitive clinical relevance, experimental validation through functional genomics assays and extended family studies is essential. These efforts could significantly contribute to a deeper understanding and accurate classification of VUS in hereditary breast cancer predispositions.