Understanding how gene-disease relationships can impact clinical utility: adaptations and challenges in hereditary cancer testing

Abstract

Background

Defining gene-disease relationships (GDRs) influences the clinical utility of hereditary cancer predisposition (HCP) multigene panel testing (MGPT) results, as variant classification relies directly on gene-disease characterization. GDR characterization for HCP is challenging due to disease prevalence, incomplete penetrance, and heterogeneity. There is insufficient data showing how gene-disease validity (GDV) scores of HCP genes affect variant classification and how GDV scores change over time. Though these issues determine the results of HCP-MGPT, their impact on short- and long-term clinical utility has not been explored in-depth.

Methods

Using an evidence-based GDV framework, genes were classified into five standardized GDV categories at the time of panel addition. We curated changes in GDV scores and classifications for HCP-MGPT over 7 years. The corresponding impact on the frequency of positive and variant of uncertain significance (VUS) results was evaluated by GDV score.

Results

Positive results were most common in Definitive evidence genes (31.5%), with none in Limited evidence genes (0%). Genes with Definitive GDRs (n = 42) remained Definitive, while most genes with Strong (6/10, 60%) and Moderate (19/24, 80%) GDRs changed categories, 8 (23.5%) of which received a clinically significant GDR downgrade. GDRs associated with low-moderate risk of breast cancer were significantly more likely to be downgraded compared to GDRs associated with rarer, high-penetrance specific phenotypes (p < 0.0001). Downgrades for all GDRs were due to new published data and updates to the GDV framework (77%, 10/13), with 23% (3/13) due to framework updates alone. Including Limited evidence genes on MGPT increased the VUS frequency by 13.7% percentage points.

Conclusions

Positive and VUS results varied by GDV category, and Limited evidence genes did not contribute to diagnostic yield. No Limited evidence genes in the category for ≥ 3 years (n = 8) were upgraded, indicating that including these genes on HCP-MGPT provides limited long-term clinical utility. Our data highlight that GDV assessment for HCP requires robust evidence and must account for variable disease penetrance and elevated prevalence in the population. Balancing the availability of a comprehensive gene menu and transparency surrounding clinical utility of novel genes will maximize identification of high-risk patients while reducing the risk of misdiagnosis through clinical false-positive results.

Supplementary Information

The online version contains supplementary material available at 10.1186/s13073-025-01499-5.

Background

The advent of next-generation sequencing (NGS) technologies has increased the rate of gene discovery over the past several years, resulting in the adoption of larger multigene panel tests (MGPT) in clinical practice. Expanded MGPT has considerably changed risk assessment and counseling surrounding genetically heterogeneous diseases such as hereditary cancer predisposition (HCP) [1]. The accessibility of such testing, along with broader recommendations for genetic testing [2], has resulted in more individuals undergoing HCP-MGPT [3, 4].

A universal challenge is determining which genes may be appropriate to test on HCP-MGPT, particularly as novel tumor predisposition gene associations are regularly published, many of which still require robust evaluation [5–7]. Understanding and interpreting such vast amounts of new genetic data requires validated, standardized approaches to determining the validity of proposed gene-disease relationships (GDRs). Establishing GDV is an essential component of selecting genes with clinical utility for HCP-MGPT. Additionally, accurate variant classification is impossible without an existing GDV curation that establishes the strength of evidence supporting a GDR. Thus, GDV plays a fundamental role in the overall accuracy and clinical utility of HCP-MGPT.

Our laboratory previously developed and published a GDV scoring system to assess the strength of GDRs, particularly in the setting of exome analysis for rare diseases [8], based in part on the framework put forth by Clinical Genome Resource (ClinGen) [9] with adaptations to allow for consistent and accurate application in a high-volume laboratory with diverse testing indications. GDV scores are categorized into 5 tiers (Definitive, Strong, Moderate, Limited, and Disputed), indicating the strength of evidence for a specific GDR. Like variant curation [10], robust GDV curation is an evolving, dynamic process; over several years, GDV frameworks have undergone revisions to reflect emerging knowledge on gene-disease curation [11]. GDV frameworks are intended to provide standardization of GDV assessments by curating evidence from published literature and establishing numerical confidence thresholds based on the quantity and quality of the evidence [8, 8, 12]. Despite efforts to standardize GDV assessments and published guidelines regarding the selection of and reporting of genes and variants on MGPT [13], gene content of MGPT can differ markedly between testing laboratories. Based on recommendations from professional guidelines [13, 14], MGPT should include only genes with established GDRs for which clinically actionable information can be obtained. However, the use of expanded HCP-MGPT has resulted in the inclusion of multiple genes with preliminary disease associations (or “candidate” genes), known as Limited evidence genes [15, 16]. Data are lacking on the short- and long-term clinical utility of including Limited evidence genes on HCP-MGPT, and while one study has reviewed GDV classification changes over time for a small subset of genes [17], this study did not evaluate genes specific to HCP and did not assess the impact on clinical utility.

We reviewed the changes of GDV scores for genes offered on a commercially available HCP-MGPT over 7 years and evaluated the effect of these changes on the clinical utility of such panels. These data are essential to inform future MGPT gene selection, define the cadence of GDV reviews, and help clinicians select MGPT that will be most informative for their patients.

Methods

Our laboratory’s database of genes associated with HCP (n = 85) was retrospectively reviewed for current GDV classifications. The total number of characterized (Moderate, Strong, and Definitive GDV categories) and uncharacterized (Limited, No Reported Evidence, and Disputed GDV categories) genes as of August 23, 2023, was recorded, and the number of genes in each tier of the GDV framework (Disputed, No Reported Evidence, Limited, Moderate, Strong, and Definitive) was calculated. GDV characterizations were based on the highest classified hereditary cancer-related GDR for each gene. Our laboratory maintains an internal database of GDV scores, data of which is shared with the Gene Curation Coalition (GenCC, https://search.thegencc.org/submitters/GENCC_000101). Additionally, variant-level classifications are shared with the ClinVar database (https://www.ncbi.nlm.nih.gov/clinvar/submitters/61756/).

Additionally, all genes offered on HCP-MGPT (n = 85) were reviewed for changes to GDV category during the time frame 2016–2023, and changes to GDV categories along with the reason for change were recorded. Review of GDV was prompted by multigene panel content updates and/or proactive literature review of publications citing new evidence for or against a gene-disease association. Changes to GDV category among different disease areas (breast cancer, other common cancers, and rare tumor types) were compared using Fisher’s exact test.

To examine how GDV scores affect variant classification, reported variants in genes on HCP-MGPTs (panels ranging between 8 and 85 HCP genes) offered at our laboratory from 2014 to 2021 were retrospectively reviewed. The frequency of pathogenic (P), likely pathogenic (LP) variants, and variants of uncertain significance (VUS) were categorized by four GDV categories: uncharacterized (No Reported Evidence, Limited, or Disputed GDV), Moderate, Strong, and Definitive based on evaluation of published clinical and experimental evidence per the current GDV framework. Frequencies were determined by reviewing the number of reported variants (VUS/LP/P) rather than the number of cases.

Finally, the overall classification of reportable variants reported on HCP-MGPT orders between July 2020 and December 2021 were retrospectively reviewed to determine how the inclusion of uncharacterized GDRs affects the overall frequency of P/LP variants and VUS in patient orders. Frequency of VUS and P/LP variants was compared between the largest possible HCP-MGPT (85 genes) with and without uncharacterized GDRs. Calculations were based on the number of reported cases.

A detailed description of our laboratory’s published GDV framework is provided in Smith et al. [8]. Revisions since initial publication reflect emerging knowledge on defining GDRs and also support alignment with the Clinical Genome Resource (ClinGen) GDV framework [9]. These include refined application of points per reported proband and a more detailed definition of GDV categories, including the assignment of negative points in the Disputed category to allow for scoring contradictory evidence against a GDR (Fig. 1, Additional File1). The updated framework assigns an individual score for each proband with a reported variant rather than pooled groupings (Fig. 1). Other updates include calibration of points based on incomplete penetrance and phenocopies seen in more common diseases. As such, the current framework designates a lower default weight for diseases exhibiting lower penetrance and/or with high heterogeneity (e.g., a patient with a common adult-onset cancer would receive fewer points compared to a rare, completely penetrant phenotype with less risk of phenocopy). A new category of evidence to incorporate case–control data was added and is now a prerequisite for the characterization of lower penetrance gene-disease relationships involving more common cancers (e.g., breast, colon) in the updated framework.

Fig. 1.

Overview of Smith, et al. (2016) framework compared to current GDV framework. Refer to Additional File 1: Table S1 for full details

Any GDR that reaches a score of Moderate, Strong, or Definitive is considered characterized with enough evidence to implicate variants as potentially causative of disease and classified as pathogenic (P) or likely pathogenic (LP). GDRs in the Limited or No Reported Evidence categories are considered uncharacterized and do not have enough evidence to associate a gene with a specific disease; therefore, variants at most can be classified as VUS [18]. Finally, GDRs in the Disputed category have evidence contradicting an association with disease and likewise cannot have any variants classified LP/P [10]. Full details of the framework are available in Additional File1. The current framework was used to evaluate all HCP GDVs in 2023.

Results

Changes in HCP GDV scores

Retrospective review of 85 HCP genes offered on MGPT as of August 2023 revealed 57.6% (n = 49) were Definitive, 14.1% (n = 12) Strong, 9.4% (n = 8) Moderate, 8.2% (n = 7) Limited, and 10.6% (n = 9) Disputed (Fig. 2).

Fig. 2.

Hereditary cancer predisposition MGPT content (n = 85) by current GDV score. Characterizations based on the highest classified gene–disease relationship for each gene

Of 85 genes, 56 (66%) did not change GDV category during the study period, while 13 (19%) were downgraded, and 16 (15%) were upgraded. The reason for downgrade was most often due to a combination of new data and GDV framework updates (n = 10; 76.9%), but almost a quarter of downgrades were due to framework updates only (n = 3; 23.1%). No downgrades were due to new data only. GDV upgrades were almost always due to new data only (n = 15; 93.8%), with one upgrade (6.3%) due to a combination of new data and framework updates. No upgrades were due to framework updates only (Fig. 3).

Fig. 3.

Rationale for changes in GDV category. GDRs were assessed for net change in category (upgrade, downgrade, or none) between 2016 and 2023. The reason for category changes were evaluated as the result of new published data, GDV framework updates, or both

GDRs in the Definitive GDV category in 2016 (n = 42) remained unchanged over the 7-year period. Most genes in the Strong (n = 6, 60%) and Moderate (n = 19, 79.2%) GDV categories at initial panel addition changed GDV score. During the assessed period, 54.3% (n = 19) of genes in the Moderate and Strong categories were upgraded to either the Definitive or Strong categories. Notably, nearly one quarter (n = 8 or 23.5%) of genes initially characterized in the Strong and Moderate categories at the time of panel addition received a clinically significant downgrade (to either Limited or Disputed), in which variants were reclassified from LP/P to VUS. In contrast, the majority (n = 5 or 62.5%) of genes in the Limited category at the time of panel addition remained in the Limited category, with only 1 gene (11.1%) upgraded to Moderate and 3 (33.3%) downgraded to the Disputed category (Fig. 4, Additional File 2: Table S2).

Fig. 4.

Changes in GDV scores over a 7-year period for all genes on HCP-MGPT (n = 85). All genes (n = 42, 100%) in the Definitive category remained Definitive with the Current GDV framework. Of 10 genes initially in the Strong category, 3 genes (30%) moved to Definitive, 2 (20%) were downgraded to Moderate, and 1 (10%) was downgraded to Disputed. Of 24 genes initially in the Moderate category, 4 genes (16.7%) moved to Definitive, 8 (33.3%) moved to Strong, 2 (8.3%) were downgraded to Limited, and 5 (20.8%) were downgraded to Disputed. Of 9 genes initially in the Limited category, 1 (11.1%) moved to Moderate, 5 (62.5%) remained Limited, and 3 (33.3%) were downgraded to Disputed. Refer to Additional File 2: Table S2 for full details

Of GDRs that received downgrades into the Disputed category (with new evidence contradicting a disease association) in the current framework (n = 9), all were previously associated with low to moderate risk breast cancer predisposition. Only one GDR (9%) associated with breast cancer was upgraded. When downgrades to GDRs over time for breast cancer (9/11) were compared to downgraded GDRs for other types of cancers (4/32), those associated with low to moderate penetrance breast cancer risk were statistically significantly more likely to be downgraded (p-value < 0.0001). Combining breast cancer GDRs with other common cancer GDRs (for colorectal cancer, gastric cancer, and prostate cancer) (9/22), this group of GDRs was also more likely to be downgraded compared to rarer tumor types (4/21), though this comparison was not statistically significant (p-value 0.19). One GDR remained the same (Strong), for BARD1-related breast cancer (Fig. 4, Additional File 2: Table S2).

No GDRs were upgraded from Limited to Strong or Definitive, and only one GDR received an upgrade from Limited to Moderate: PDGFRA, which is associated with a rare hereditary gastrointestinal stromal tumor predisposition syndrome. GDRs that were upgraded to a stronger GDV category were more likely to be associated with rarer tumor types (11/21) such as brain cancer, endocrine tumors, renal cancers, or melanoma than common tumor types (5/22; p-value 0.06). For full details on all 85 genes, see Additional File 2: Table S2.

Table 1 highlights noteworthy GDV upgrades and downgrades in a subset of genes (n = 18) along with reasons for changes. GDRs that received clinically significant downgrades into the new Disputed category (n = 9) were due to a combination of assigning a lower default score per proband (to account for phenocopy and heterogeneity in lower penetrance HCP genes) and new case–control data showing a lack of association with the curated disease(s). Ten additional GDRs resulted in GDV category upgrades. These 10 GDRs received higher scores due to additional case reports, cohort studies, new functional data, or new animal model data strengthening the evidence supporting the GDRs.

Table 1.

Selection of genes with GDV category changes over the period assessed

Disease area	Gene symbol	Classification at time of panel addition (2016 Framework)	Reason for change
			Current classification	Cohort studies	Case report(s)	Co-segregation data	Functional data	Case–control data	New animal model
Breast cancer	ATM	Strong	Definitive	+				+
	BLM	Strong	Disputed					+
	FAM175A	Moderate	Disputed					+
	FANCC	Moderate	Disputed					+
	MRE11A	Moderate	Disputed					+
	NBN	Moderate	Disputed					+
	RAD50	Moderate	Disputed					+
	RECQL	Limited	Disputed					+
	XRCC2	Moderate	Disputed					+
Colorectal cancer/polyposis	MSH3	Moderate	Strong		+	+
	NTHL1	Strong	Definitive		+		+
Other (rare and/or specific tumor spectrum*)	MAX	Moderate	Definitive	+	+	+
	MET	Moderate	Definitive		+		+
	PDGFRA	Limited	Moderate		+		+		+
	PHOX2B	Moderate	Strong		+		+
	POT1	Moderate	Strong		+		+
	SMARCA4	Moderate	Strong	+	+
	SUFU	Strong	Definitive		+		+

^*MAX, paraganglioma/pheochromocytoma; MET, papillary renal cancer; PDGFRA, gastrointestinal stromal tumor; PHOX2B, neuroblastoma; POT1, melanoma and sarcoma; SMARCA4, atypical teratoid/rhabdoid tumors and small cell carcinoma of the ovary, hypercalcemic type; SUFU, SUFU-related disorders (basal cell carcinoma, medulloblastoma, meningioma, and gonadal tumors). Refer to Additional File 2: Table S2 for full details

Effect of GDV scores on variant classification and diagnostic yield

Of reportable variants during the study period, the frequency of P/LP variants was 31.5%, 19.9%, 11.0%, and 0% among genes in the Definitive, Strong, Moderate, and Limited/Disputed categories, respectively. The frequency of VUS was inversely proportional to the frequency of P/LP variants and increased as GDV scores decreased (68.5%, 80.1%, 89.0%, and 100% for the Definitive, Strong, Moderate, and uncharacterized categories, respectively). No variants in Limited/Disputed genes were classified as P/LP (Fig. 5), per ACMG classification guidelines.

Fig. 5.

Reportable variant classification frequencies by GDV score. *Uncharacterized GDRs include both Limited and Disputed GDRs. Dark blue bars indicate the percentage of variants reported that were classified as pathogenic (P) or likely pathogenic (LP). Light blue bars indicate the percentage of variants reported that were classified as VUS. Higher GDV scores resulted in more P/LP variants reported. GDRs with Limited GDV had no variants classified as LP/P. Variants classified as benign and likely benign are not shown

The frequency of VUS resulted from 22,558 HCP-MGPT orders that included genes with uncharacterized GDRs was 54.7% (12,342 VUS). Removing genes with uncharacterized GDRs from this analysis yielded a frequency of 41.0% (9257 VUS reported), corresponding to a 13.7% percentage point reduction in VUS. Conversely, the frequency of LP/P variants reported on HCP-MGPT orders with and without uncharacterized GDRs remained the same at 12.4% (Fig. 6).

Fig. 6.

Frequency of P/LP variants and VUS on HCP-MGPT orders (85–91 genes) with and without uncharacterized GDRs. The dark blue solid and textured bars indicate the frequency of P/LP variants reported on HCP-MGPT (85–91 genes) with and without uncharacterized GDRs, respectively. The light blue solid and textured bars indicate the frequency of VUS reported on HCP-MGPT with and without uncharacterized GDRs, respectively. The frequency of P/LP was unchanged with the inclusion of uncharacterized GDRs. The frequency of overall VUS reported increased nearly 14% with the inclusion of uncharacterized GDRs

Discussion

GDV score changes

Our analysis of how HCP-GDV scores changed over 7 years revealed important trends: first, GDRs with Definitive GDV remained unchanged, which indicates that once GDRs reach this GDV category, a frequent cadence of reassessment is not as crucial as it is for GDRs in the Moderate or Strong categories, the majority of which changed GDV classifications during the assessed period (Fig. 3). GDRs that moved from the Moderate-to-Strong or Definitive or Strong-to-Definitive categories (n = 15) did so due to newly published case reports and cohort studies detailing the phenotypes and variants in more individuals. The GDRs that resulted in stronger GDV scores were largely higher-penetrance tumor syndromes with highly specific and/or rare phenotypes (13/15, 86.7%) (Table 1, Additional File 2: Table S2). Moderate evidence GDRs that were downgraded were predominantly in genes associated with lower penetrance for common tumor types such as breast cancer (5/7, 71.4%). GDRs in the Definitive and Limited categories were unlikely to undergo reclassification, whereas Strong and Moderate categories were more influenced by newly available evidence and adaptations to the framework (Fig. 3). Due to the changes observed in both Moderate and Strong GDVs during the assessed period, frequent reassessment at least every 1–2 years is crucial to maintaining an up-to-date gene curation database, data which agrees with a previous publication reviewing the progression of GDVs [17]. Importantly, nearly a quarter (23.5%) of initially characterized genes were downgraded to uncharacterized categories (Limited or Disputed), all of which were thought to be associated with low to moderate penetrance breast cancer predisposition (Table 1, Additional File 2: Table S2). These changes were clinically significant, as all variants initially classified as LP or P for these GDRs were downgraded to VUS due to the change from a characterized to an uncharacterized GDR.

Figure 3 outlines reasons for GDV category upgrades or downgrades based on new data and/or updates to the GDV framework. Notably, no GDR upgrades (0/16) were the result of framework updates only, and always relied on new published data. In contrast, GDR downgrades most often occurred due to a combination of new data and framework updates (10/13; 76.9%), reflecting both the calibration for proband-variant scoring for diseases with high heterogeneity and incomplete penetrance and the incorporation of negative points scoring for case–control analyses showing lack of an association with disease. Nearly a quarter of downgrades (3/13; 23.1%) were due to framework updates alone.

A challenging aspect of GDV scoring for higher prevalence phenotypes like breast cancer is possible premature characterization of genes; this can lead to ascribing “pathogenic” variants for a gene-disease relationship (GDR) that may ultimately be disputed. An example of this scenario is evident in the initially presumed associations of NBN and RECQL heterozygotes with an increased risk for breast cancer, associations which have since been refuted [19–24]. Initial designation of these genes as associated with breast cancer came primarily from cohort studies or case series identifying multiple rare and/or premature truncating variants in individuals with breast cancer [25–28], but also from case–control studies in smaller populations and/or assessing only single variants [19, 29–31]. Notably, the NBN gene was previously included in professional screening and management recommendations for hereditary breast cancer [32] and has now been removed. The early studies showing unique variants in probands from breast cancer cohorts were instead incidental findings, and smaller or single-variant case–control studies were underpowered to detect accurate associations. The downgrades were due to a combination of individual patient-variant scoring re-calibration (scores down-weighted) to account for phenocopy/incomplete penetrance and new literature in the form of robust, highly powered case–control studies that showed no association with breast cancer [33, 34].

Even with published gene curation guidelines [9], curation of genes associated with hereditary cancer (particularly low to moderate penetrance for common cancers such as breast or colon) predisposition remains challenging, as evidenced by these examples [35]. Common diseases like cancer are, by definition, seen in a higher percentage of the population and can be caused by a combination of factors, including lifestyle, environment, and polygenic influences. This complexity leads to potential interference based on incomplete penetrance and phenocopies. Rare variant detection in individuals with common cancers such as breast or colon therefore cannot carry the same weight as rare variants seen in individuals with highly penetrant, extremely low prevalence diseases (i.e., rare diseases), and gene-disease curation frameworks require adaptation to take this into account [19–22]. These downgrade examples highlight the known challenges and complexities involved in GDV assessment for hereditary cancer [36] and can serve as a cautionary tale in the addition of such genes to HCP-MGPT without adequate data accumulation. Reclassifications such as these also pose significant challenges for clinical management and interpretation [37], a burden which is noteworthy on multiple levels: clinician follow-up requires substantial time and effort, patients who may have previously received positive results for MGPT for a re-classified gene will now receive an uncertain or negative result (contributing to the psychosocial burden), and diagnostic laboratories must re-assess the placement of such genes on panels [38–40]. Re-classification of GDRs therefore impact a larger number of downstream components compared to single variant-level re-classifications, as all patients with variants reported in such genes will be impacted. Of important note, we expect that the use of the more conservative, re-calibrated patient-variant scoring for lower penetrance hereditary cancer associations and refined case–control scoring in the current GDV framework will prevent future premature characterization of these GDRs and serve as solutions to the challenges faced. Lastly, these examples also highlight the importance of continued review and revision of GDV curation frameworks as the amount of available genetic data continues to rise and knowledge of gene-disease relationships evolves.

Limited evidence genes and clinical utility

In recent years, including Limited evidence genes on clinical HCP-MGPT has become mainstream. These large panels are attractive to clinicians and patients with the expectation that they will detect more pathogenic variants (and increase diagnostic yield) in individuals undergoing testing, and they can help reduce costs by ordering only one large panel rather than ordering reflex/update testing, which can become cost prohibitive. Large HCP-MGPT panels can also have the undesirable effect of increasing the frequency of VUS, which can create considerable uncertainty with clinical management for both clinicians and patients [12, 41].

Because no variants in Limited/Disputed genes are classified as LP/P, our data show that the inclusion of genes with Limited evidence on HCP-MGPT does not increase diagnostic yield or clinical utility in the short term (Figs. 4 and 5). Limited evidence genes do not have sufficient evidence to assume a clinically meaningful GDR. Therefore, the impact of any variant in a Limited evidence gene will be at most a VUS, and Limited evidence genes should be designated as such in clinical test reports per professional guidelines [10, 13]. Notably, comparison of HCP-MGPT orders with and without uncharacterized genes revealed that the inclusion of uncharacterized genes results in a nearly 14% percentage point increase in the VUS frequency (54.7% versus 41.0%).

Based on observations in the rare disease setting in which Limited evidence genes are often upgraded as evidence accumulates over time [42], it is plausible to suggest including Limited evidence genes on HCP-MGPT. However, since the initial HCP gene discovery “boom” of the 1990s, and the additional discoveries leading to MGPT in the early 2000s to 2010s, the rate of gene discovery and characterization in the hereditary cancer setting have proven to be significantly less robust even since the advent of whole exome and genome sequencing [43–45], trends of which are supported by our data. Of 9 cancer genes with Limited evidence at the time of panel addition during our assessed 7-year period, only 1 (11.1%), PDGFRA, accrued enough evidence for a stronger GDV classification (Moderate) due to additional published case reports and cohort studies, and was characterized shortly after initial assessment (within 18 months). The remaining 7 (88%) Limited evidence genes remained uncharacterized (or were downgraded to Disputed) during the assessed period (Fig. 3). This data is largely in agreement with another published assessment of GDV changes over time, which shows that Limited evidence genes that have remained in that category for more than 3 years largely remain uncharacterized or are downgraded to Disputed or Refuted [18]. Limited evidence genes that ultimately do become characterized tend to be in more specific and/or rarer cancer phenotypes, and also tend to do so within 1 to 3 years of initial publication. Our data show that the longer-term clinical utility of including HCP genes in the Limited category that have been Limited for more than 3 years is therefore negligible. These data have important implications for HCP-MGPT panel design, suggesting that newly published genes in the setting of HCP, particularly for common tumors such as breast and colon cancers, would benefit from the accrual of more robust data (particularly in the form of large case–control studies, or case series with reference to a control population) before consideration of addition to clinically available MGPT. Re-curation at least every 1–2 years would be essential, and if such data has not been amassed within 3–5 years of initial publication, it is highly unlikely that the GDR would become characterized in the future. Acknowledging that published literature is often biased toward positive association data, the accrual of data that does not support a gene-disease relationship also has value. As such, data sharing and collaboration are a large driving factor in the eventual resolution of Limited and Moderate evidence gene-disease relationships.

These findings should be interpreted in the context of several caveats and limitations. GDV curation, despite using an established framework, is inherently subjective by nature and can result in differing scores dependent on individual curators and is also subject to any ascertainment or other biases present in the scored published literature. In addition, the trends in GDV classification changes we report here are limited to specific genes on a curated multigene HCP test during the study period, which was designed to maximize clinical utility; therefore, the total number of Limited evidence genes evaluated does not reflect all potential limited evidence HCP genes that may have been reported in published literature. More data are necessary to determine trends of how GDV scores change over longer periods, particularly for newly implicated HCP genes. GDR scores reported in this manuscript are from the end of the study period, August 2023, and may be subject to change pending new data evaluation. The most up-to-date classifications can be found in the publicly available Gene Curation Coalition (https://thegencc.org/).

Since this study was limited to evaluating the impact that GDR recharacterization has on clinical utility, it does not address all the possible reasons for specific variant classifications or re-classifications beyond the scope of gene curation. For example, newly published functional assays, understanding of mechanism of disease/mutational mechanisms, and better interpretation of RNA data are a few examples of data that contribute to re-classification of variants. An exploration of all the specific parameters that may affect variant classification is out of scope for this study but would be a valuable addition to the literature.

GDV classifications reported here for the 85 HCP genes were based on the highest-characterized GDR for cancer predisposition. In some cases, the genes analyzed may have allelic disease associations. For example, SMARCA4 was curated for rhabdoid tumor predisposition syndrome and not for Coffin–Siris syndrome as part of this analysis. Providing curation data for all potential associated tumors and allelic diseases was out of scope for this analysis but is an important consideration when assessing GDRs for HCP genes to help guide screening and clinical management for potential multi-tumor risks.

Conclusions

Reviewing GDV classification changes over a 7-year period highlights the importance of accurate and ongoing GDR assessment, especially for common cancer phenotypes such as breast cancer. When newly implicated cancer susceptibility genes are identified via NGS technologies, it is essential to assess these GDRs using a variety of lines of evidence, including reproducible case–control data, to account for general population prevalence and heterogeneity; this can avoid premature characterization of genes before a true association with hereditary cancer has been established. Importantly, our data showed that several genes that initially reached Moderate GDV in 2016 were eventually downgraded. This stresses the need for continued reassessment (at least every 1–2 years) of Limited and newly characterized Moderate GDRs as new data becomes available. Additionally, including Limited evidence genes on HCP-MGPT significantly increases the VUS frequency and does not contribute to clinical utility or diagnostic yield in the short term. Importantly, we found that over longer periods, characterization of Limited evidence genes in the setting of low to moderate penetrance for common disease such as breast cancer is exceedingly rare. Therefore, our study data indicates that the risk of missing clinically significant results by omitting Limited evidence genes on HCP-MGPT is negligible. This study highlights the benefits of dynamic, data-driven GDV curation that can adapt to emerging knowledge of gene-disease relationships. Robust GDV curation guides HCP-MGPT panel design to maximize the identification of hereditary cancer patients while minimizing clinical false-positive results on HCP-MGPT, directly informing the utility of such panels in clinical practice.

Abbreviations

GDV

Gene-disease validity

GDR

Gene-disease relationship

HCP

Hereditary cancer predisposition

MGPT

Multigene panel testing

VUS

Variant of uncertain significance

LP

Likely pathogenic variant

P

Pathogenic variant

Glossary

Gene-disease relationship (GDR)

describes the connection between a gene and a disease, regardless of the strength of evidence supporting it.

Gene-disease validity (GDV)

refers to the strength of evidence supporting a link between a specific gene and a particular disease.

Characterized GDR

A GDR with enough evidence to implicate variants as disease-causing (encompasses GDRs with Moderate, Strong, and Definitive GDV scores).

Uncharacterized GDR

A GDR that does not have enough evidence to support a link between variants and a given disease (encompasses Limited evidence and No Reported Evidence GDV scores).

Disputed GDR

A GDR with evidence contradicting the role of a gene in a particular disease.

Authors’ contributions

Conceptualization: J.H-M., B.W., M.T., A.W., J.M.H. Data Curation: J.H-M., M.T., C.H., B.W., J.M.H., A.W. Formal Analysis: J.H-M., M.T., C.H. Methodology: J.H-M., M.T., B.W., A.W., J.M.H., C.H. Visualization: J.H-M., M.T. Writing-original draft: J.H-M. Writing-reviewing and editing: J.H-M., A.W., J.M.H., B.W., M.T., C.H., S.H., G.V. All authors read and approved the final manuscript.

Funding

The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: The authors received no financial support for the research, authorship, and/or publication of this article. All authors are full-time, salaried employees of Ambry Genetics.

Data availability

Availability of data and materials All data generated or analyzed during this study are included in this published article and its additional files. Ambry Genetics supports open data sharing. Internal GDV scores reported in this manuscript are submitted to the Gene Curation Coalition Database (GenCC DB; https://thegencc.org/), and all variants identified on clinical testing are deposited into ClinVar (https://www.ncbi.nlm.nih.gov/clinvar/).

Declarations

Ethics approval and consent to participate.

WCG Institutional Review Board determined the study to be exempt from the Office for Human Research Protections Regulations for the Protection of Human Subjects (45 CFR 46) under category 4. Retrospective data analysis of de-identified data exempted the study from the requirement to receive consent from patients. We confirm that this research conforms to the principles of the Helsinki Declaration.

Consent for publication

Not applicable.

Competing interests

All authors are employees of Ambry Genetics Corporation at the time of submission.