Identifying rare,medically relevant variation via population-based genomic screening in Alabama: opportunities and pitfalls

PurposeTo evaluate the effectiveness and specificity of population-based genomic screening in Alabama.MethodsThe Alabama Genomic Health Initiative (AGHI) has enrolled and evaluated 5369 participants for the presence of pathogenic/likely pathogenic (P/LP) variants using the Illumina Global Screening Array (GSA), with validation of all P/LP variants via Sanger sequencing in a CLIA-certified laboratory before return of results.ResultsAmong 131 variants identified by the GSA that were evaluated by Sanger sequencing, 67 (51%) were false positives (FP). For 39 of the 67 FP variants, a benign/likely benign variant was present at or near the targeted P/LP variant. Variants detected within African American individuals were significantly enriched for FPs, likely due to a higher rate of nontargeted alternative alleles close to array-targeted P/LP variants.ConclusionIn AGHI, we have implemented an array-based process to screen for highly penetrant genetic variants in actionable disease genes. We demonstrate the need for clinical validation of array-identified variants in direct-to-consumer or population testing, especially for diverse populations.     

The ACMG SF v3.0 gene list increases returnable variant detection by 22% when compared with v2.0 in the ClinSeq cohort

PurposeThe American College of Medical Genetics and Genomics (ACMG) recommends the return of pathogenic and likely pathogenic (P/LP) secondary findings from exome and genome sequencing. The latest version (ACMG secondary finding [SF] v3.0) includes 14 additional genes. We interrogated the ClinSeq cohort for variants in these genes to determine the additional yield in unselected individuals.MethodsExome data from 1473 individuals (60% White, 34% African American or Black, 6% other) were analyzed. We restricted our analyses to coding variants; +1,+2,–1, and –2 splice site variants; and the pathogenic GAA variant, NM_000152.5:c.-32-13T>G. Variants were assessed with slightly modified ACMG/Association of Molecular Pathology guidelines.ResultsA total of 25 P/LP variants were identified. In total, 7 individuals had P/LP variants in genes recommended for return of heterozygous variants, namely HNF1A (1), PALB2 (3), TMEM127 (1), and TTN (2). In total, 4 individuals had a homozygous variant in a gene recommended for biallelic variant return, namely HFE, NM_000410.3(HFE):c.845G>A p.Cys282Tyr. A total of 17 P/LP variants were identified in the heterozygous state in genes recommended only for biallelic variant reporting and were not returned. The frequency of returnable P/LP variants did not significantly differ by race.ConclusionUsing the ACMG SF v3.0, the returnable P/LP variant frequency increased in the ClinSeq cohort by 22%, from 3.4% (n = 50, ACMG SF v2.0) to 4.1% (n = 61, ACMG SF v3.0).     

Investigation of de novo variation in pediatric cardiomyopathy

Pediatric cardiomyopathies can be caused by variants in genes encoding the sarcomere and cytoskeleton in cardiomyocytes. Variants are typically inherited in an autosomal dominant manner with variable expressivity. De novo variants have been reported, however their overall frequency is largely unknown. We sought to determine the rate of de novo, pathogenic and likely pathogenic (P/LP) variants in children with a diagnosis of hypertrophic, dilated, or restrictive cardiomyopathy (HCM, DCM, or RCM), and to compare disease outcomes between individuals with and without a de novo variant. A retrospective record review identified 126 individuals with HCM (55%), DCM (37%), or RCM (8%) ≤18 years of age who had genetic testing. Overall, 50 (40%) had positive genetic testing and 18% of P/LP variants occurred de novo. The rate of de novo variation in those with RCM (80%) was higher than in those with HCM (9%) or DCM (20%). There was evidence of germline mosaicism in one family with RCM. Individuals with de novo variants were more likely than those without to have a history of arrhythmia (p = .049), sudden cardiac arrest (p = .024), hospitalization (p = .041), and cardiac transplantation (p = .030). The likelihood of de novo variation and impact on family risk and screening should be integrated into genetic counseling.     

Successful application of genome sequencing in a diagnostic setting: 1007 index cases from a clinically heterogeneous cohort

Despite clear technical superiority of genome sequencing (GS) over other diagnostic methods such as exome sequencing (ES), few studies are available regarding the advantages of its clinical application. We analyzed 1007 consecutive index cases for whom GS was performed in a diagnostic setting over a 2-year period. We reported pathogenic and likely pathogenic (P/LP) variants that explain the patients’ phenotype in 212 of the 1007 cases (21.1%). In 245 additional cases (24.3%), a variant of unknown significance (VUS) related to the phenotype was reported. We especially investigated patients which had had ES with no genetic diagnosis (n = 358). For this group, GS diagnostic yield was 14.5% (52 patients with P/LP out of 358). GS should be especially indicated for ES-negative cases since up to 29.6% of them  could benefit from GS testing (14.5% with P/LP, n = 52 and 15.1% with VUS, n = 54). Genetic diagnoses in most of the ES-negative/GS-positive cases were determined by technical superiority of GS, i.e., access to noncoding regions and more uniform coverage. Importantly, we reported 79 noncoding variants, of which, 41 variants were classified as P/LP. Interpretation of noncoding variants remains challenging, and in many cases, complementary methods based on direct enzyme assessment, biomarker testing and RNA analysis are needed for variant classification and diagnosis. We present the largest cohort of patients with GS performed in a clinical setting to date. The results of this study should direct the decision for GS as standard second-line, or even first-line stand-alone test.Subject terms: Genetics research, Genomics     

Targeted genetic analysis in a large cohort of familial and sporadic cases of aneurysm or dissection of the thoracic aorta

PurposeThoracic aortic aneurysm/aortic dissection (TAAD) is a disorder with highly variable age of onset and phenotype. We sought to determine the prevalence of pathogenic variants in TAAD-associated genes in a mixed cohort of sporadic and familial TAAD patients and identify relevant genotype–phenotype relationships.MethodsWe used a targeted polymerase chain reaction and next-generation sequencing–based panel for genetic analysis of 15 TAAD-associated genes in 1,025 unrelated TAAD cases.ResultsWe identified 49 pathogenic or likely pathogenic (P/LP) variants in 47 cases (4.9% of those successfully sequenced). Almost half of the variants were in nonsyndromic cases with no known family history of aortic disease. Twenty-five variants were within FBN1 and two patients were found to harbor two P/LP variants. Presence of a related syndrome, younger age at presentation, family history of aortic disease, and involvement of the ascending aorta increased the risk of carrying a P/LP variant.ConclusionGiven the poor prognosis of TAAD that is undiagnosed prior to acute rupture or dissection, genetic analysis of both familial and sporadic cases of TAAD will lead to new diagnoses, more informed management, and possibly reduced mortality through earlier, preclinical diagnosis in genetically determined cases and their family members.     

Genome sequencing as a first-line genetic test in familial dilated cardiomyopathy

PurposeWe evaluated genome sequencing (GS) as an alternative to multigene panel sequencing (PS) for genetic testing in dilated cardiomyopathy (DCM).MethodsForty-two patients with familial DCM underwent PS and GS, and detection rates of rare single-nucleotide variants and small insertions/deletions in panel genes were compared. Loss-of-function variants in 406 cardiac-enriched genes were evaluated, and an assessment of structural variation was performed.ResultsGS provided broader and more uniform coverage than PS, with high concordance for rare variant detection in panel genes. GS identified all PS-identified pathogenic or likely pathogenic variants as well as two additional likely pathogenic variants: one was missed by PS due to low coverage, the other was a known disease-causing variant in a gene not included on the panel. No loss-of-function variants in the extended gene set met clinical criteria for pathogenicity. One BAG3 structural variant was classified as pathogenic.ConclusionOur data support the use of GS for genetic testing in DCM, with high variant detection accuracy and a capacity to identify structural variants. GS provides an opportunity to go beyond suites of established disease genes, but the incremental yield of clinically actionable variants is limited by a paucity of genetic and functional evidence for DCM association.     

The Clinical Variant Analysis Tool: Analyzing the evidence supporting reported genomic variation in clinical practice

PurposeGenomic test results, regardless of laboratory variant classification, require clinical practitioners to judge the applicability of a variant for medical decisions. Teaching and standardizing clinical interpretation of genomic variation calls for a methodology or tool.MethodsTo generate such a tool, we distilled the Clinical Genome Resource framework of causality and the American College of Medical Genetics/Association of Molecular Pathology and Quest Diagnostic Laboratory scoring of variant deleteriousness into the Clinical Variant Analysis Tool (CVAT). Applying this to 289 clinical exome reports, we compared the performance of junior practitioners with that of experienced medical geneticists and assessed the utility of reported variants.ResultsCVAT enabled performance comparable to that of experienced medical geneticists. In total, 124 of 289 (42.9%) exome reports and 146 of 382 (38.2%) reported variants supported a diagnosis. Overall, 10.5% (1 pathogenic [P] or likely pathogenic [LP] variant and 39 variants of uncertain significance [VUS]) of variants were reported in genes without established disease association; 20.2% (23 P/LP and 54 VUS) were in genes without sufficient phenotypic concordance; 7.3% (15 P/LP and 13 VUS) conflicted with the known molecular disease mechanism; and 24% (91 VUS) had insufficient evidence for deleteriousness.ConclusionImplementation of CVAT standardized clinical interpretation of genomic variation and emphasized the need for collaborative and transparent reporting of genomic variation.     

Diagnostic yield and clinical utility of whole exome sequencing using an automated variant prioritization system,EVIDENCE

EVIDENCE, an automated variant prioritization system, has been developed to facilitate whole exome sequencing analyses. This study investigated the diagnostic yield of EVIDENCE in patients with suspected genetic disorders. DNA from 330 probands (age range, 0-68 years) with suspected genetic disorders were subjected to whole exome sequencing. Candidate variants were identified by EVIDENCE and confirmed by testing family members and/or clinical reassessments. EVIDENCE reported a total 228 variants in 200 (60.6%) of the 330 probands. The average number of organs involved per patient was 4.5 ± 5.0. After clinical reassessment and/or family member testing, 167 variants were identified in 141 probands (42.7%), including 105 novel variants. These variants were confirmed as being responsible for 121 genetic disorders. A total of 103 (61.7%) of the 167 variants in 95 patients were classified as pathogenic or probably to be pathogenic before, and 161 (96.4%) variants in 137 patients (41.5%) after, clinical assessment and/or family member testing. Factor associated with a variant being regarded as causative includes similar symptom scores of a gene variant to the phenotype of the patient. This new, automated variant interpretation system facilitated the diagnosis of various genetic diseases with a 42.7% diagnostic yield.     

Pathogenic germline variants in patients with features of hereditary renal cell carcinoma: Evidence for further locus heterogeneity

Inherited renal cell carcinoma (RCC) is associated with multiple familial cancer syndromes but most individuals with features of non‐syndromic inherited RCC do not harbor variants in the most commonly tested renal cancer predisposition genes (CPGs). We investigated whether undiagnosed cases might harbor mutations in CPGs that are not routinely tested for by testing 118 individuals with features suggestive of inherited RCC (family history of RCC, two or more primary RCC aged <60 years, or early onset RCC ≤46 years) for the presence of pathogenic variants in a large panel of CPGs. All individuals had been prescreened for pathogenic variants in the major RCC genes. We detected pathogenic or likely pathogenic (P/LP) variants of potential clinical relevance in 16.1% (19/118) of individuals, including P/LP variants in BRIP1 (n = 4), CHEK2 (n = 3), MITF (n = 1), and BRCA1 (n = 1). Though the power to detect rare variants was limited by sample size the frequency of truncating variants in BRIP1, 4/118, was significantly higher than in controls (P = 5.92E‐03). These findings suggest that the application of genetic testing for larger inherited cancer gene panels in patients with indicators of a potential inherited RCC can increase the diagnostic yield for P/LP variants. However, the clinical utility of such a diagnostic strategy requires validation and further evaluation and in particular, confirmation of rarer RCC genotype‐phenotype associations is required.     

Cancer Risk C (CR-C), a functional genomics test is a sensitive and rapid test for germline mismatch repair deficiency

PurposeHeritable pathogenic variants in the DNA mismatch repair (MMR) pathway cause Lynch syndrome, a condition that significantly increases risk of colorectal and other cancers. At least half of individuals tested using gene panel sequencing have a variant of uncertain significance or no variant identified leading to no diagnosis. To fill this diagnostic gap, we developed Cancer Risk C (CR-C), a flow variant assay test.MethodsIn response to treatment with an alkylating agent, individual assays of the nuclear translocation of MLH1, MSH2, BARD1, PMS2, and BRCA2 proteins and the nuclear phosphorylation of the ATM and ATR proteins distinguished pathogenic/likely pathogenic (P/LP) from benign/likely benign variants in MMR genes.ResultsA risk classification score based on MLH1, MSH2, and ATR assays was 100% sensitive and 98% specific. Causality of MMR P/LP variants was shown through gene editing and rescue. In individuals with suspected Lynch syndrome but no P/LP, CR-C identified most (73%) as having germline MMR defects. Direct comparison of CR-C on matched blood samples and lymphoblastoid cell lines yielded comparable results (r² > 0.9).ConclusionFor identifying germline MMR defects, CR-C provides augmentation to traditional panel sequencing through greater accuracy, shorter turnaround time (48 hours), and performance on blood with minimal sample handling.     

Phenotypic expansion illuminates multilocus pathogenic variation

PurposeMultilocus variation—pathogenic variants in two or more disease genes—can potentially explain the underlying genetic basis for apparent phenotypic expansion in cases for which the observed clinical features extend beyond those reported in association with a “known” disease gene.MethodsAnalyses focused on 106 patients, 19 for whom apparent phenotypic expansion was previously attributed to variation at known disease genes. We performed a retrospective computational reanalysis of whole-exome sequencing data using stringent Variant Call File filtering criteria to determine whether molecular diagnoses involving additional disease loci might explain the observed expanded phenotypes.ResultsMultilocus variation was identified in 31.6% (6/19) of families with phenotypic expansion and 2.3% (2/87) without phenotypic expansion. Intrafamilial clinical variability within two families was explained by multilocus variation identified in the more severely affected sibling.ConclusionOur findings underscore the role of multiple rare variants at different loci in the etiology of genetically and clinically heterogeneous cohorts. Intrafamilial phenotypic and genotypic variability allowed a dissection of genotype–phenotype relationships in two families. Our data emphasize the critical role of the clinician in diagnostic genomic analyses and demonstrate that apparent phenotypic expansion may represent blended phenotypes resulting from pathogenic variation at more than one locus.     

The genetic landscape of polycystic kidney disease in Ireland

Polycystic kidney diseases (PKDs) comprise the most common Mendelian forms of renal disease. It is characterised by the development of fluid-filled renal cysts, causing progressive loss of kidney function, culminating in the need for renal replacement therapy or kidney transplant. Ireland represents a valuable region for the genetic study of PKD, as family sizes are traditionally large and the population relatively homogenous. Studying a cohort of 169 patients, we describe the genetic landscape of PKD in Ireland for the first time, compare the clinical features of patients with and without a molecular diagnosis and correlate disease severity with autosomal dominant pathogenic variant type. Using a combination of molecular genetic tools, including targeted next-generation sequencing, we report diagnostic rates of 71–83% in Irish PKD patients, depending on which variant classification guidelines are used (ACMG or Mayo clinic respectively). We have catalogued a spectrum of Irish autosomal dominant PKD pathogenic variants including 36 novel variants. We illustrate how apparently unrelated individuals carrying the same autosomal dominant pathogenic variant are highly likely to have inherited that variant from a common ancestor. We highlight issues surrounding the implementation of the ACMG guidelines for variant pathogenicity interpretation in PKD, which have important implications for clinical genetics.Subject terms: Polycystic kidney disease, Disease genetics, Next-generation sequencing, Genetics research, Genetic testing     

The rate of secondary genomic findings in the Saudi population

Secondary findings (SF) are defined as genetic conditions discovered unintentionally during an evaluation of raw data for another disease. We aimed to identify the rate of secondary genetic findings in the Saudi population in the 59 genes of the American College of Medical Genetics and Genomics (ACMG) list. In our study, the raw data of 1254 individuals, generated from exome sequencing for clinical purposes, were studied. Variants detected in the 59 genes on the ACMG list of secondary findings were investigated. Pathogenicity classifications were assigned to those variants based on the ACMG scoring system. We identified 2409 variants in the 59 gene list, 45 variants were classified as pathogenic/likely pathogenic variants according to the ACMG classification. The LDLR gene had the greatest number of pathogenic/likely pathogenic variants 12%. Cardiovascular genetic diseases had the highest frequency of disorders detected as secondary findings. In this study, the overall rate of positive cases identified with secondary findings in the Saudi population was 8%. The different in our current study and the previous studies in Saudi Arabia can be explained by the differences between the sequencing method, the criteria used for variant classification, the availability of newer evidence at the time of the publication, and the fact that we identified Saudi novel variants never reported in other populations.     

Clinical and laboratory interpretation of mitochondrial mRNA variants

Interpretation of mitochondrial protein‐encoding (mt‐mRNA) variants has been challenging due to mitochondrial characteristics that have not been addressed by American College of Medical Genetics and Genomics guidelines. We developed criteria for the interpretation of mt‐mRNA variants via literature review of reported variants, tested and refined these criteria by using our new cases, followed by interpreting 421 novel variants in our clinical database using these verified criteria. A total of 32 of 56 previously reported pathogenic (P) variants had convincing evidence for pathogenicity. These variants are either null variants, well‐known disease‐causing variants, or have robust functional data or strong phenotypic correlation with heteroplasmy levels. Based on our criteria, 65.7% (730/1,111) of variants of unknown significance (VUS) were reclassified as benign (B) or likely benign (LB), and one variant was scored as likely pathogenic (LP). Furthermore, using our criteria we classified 2, 12, and 23 as P, LP, and LB, respectively, among 421 novel variants. The remaining stayed as VUS (91.2%). Appropriate interpretation of mt‐mRNA variants is the basis for clinical diagnosis and genetic counseling. Mutation type, heteroplasmy levels in different tissues of the probands and matrilineal relatives, in silico predictions, population data, as well as functional studies are key points for pathogenicity assessments.     

Genetic drivers of Cushing’s disease: Frequency and associated phenotypes

PurposeCushing’s disease (CD) is often explained by a single somatic sequence change. Germline defects, however, often go unrecognized. We aimed to determine the frequency and associated phenotypes of genetic drivers of CD in a large cohort.MethodsWe studied 245 unrelated patients with CD (139 female, 56.7%), including 230 (93.9%) pediatric and 15 (6.1%) adult patients. Germline exome sequencing was performed in 184 patients; tumor exome sequencing was also done in 27 of them. A total of 43 germline samples and 92 tumor samples underwent Sanger sequencing of specific genes. Rare variants of uncertain significance, likely pathogenic (LP), or pathogenic variants in CD-associated genes, were identified.ResultsGermline variants (13 variants of uncertain significance, 8 LP, and 11 pathogenic) were found in 8 of 19 patients (42.1%) with positive family history and in 23 of 226 sporadic patients (10.2%). Somatic variants (1 LP and 7 pathogenic) were found in 20 of 119 tested individuals (16.8%); one of them had a coexistent germline defect. Altogether, variants of interest were identified at the germline level in 12.2% of patients, at the somatic level in 7.8%, and coexisting germline and somatic variants in 0.4%, accounting for one-fifth of the cohort.ConclusionWe report an estimate of the contribution of multiple germline and somatic genetic defects underlying CD in a single cohort.     

Next-generation sequencing of 32 genes associated with hereditary aortopathies and related disorders of connective tissue in a cohort of 199 patients

PurposeHeritable factors play an important etiologic role in connective tissue disorders (CTD) with vascular involvement, and a genetic diagnosis is getting increasingly important for gene-tailored, personalized patient management.MethodsWe analyzed 32 disease-associated genes by using targeted next-generation sequencing and exome sequencing in a clinically relevant cohort of 199 individuals. We classified and refined sequence variants according to their likelihood for pathogenicity.ResultsWe identified 1 pathogenic variant (PV; in FBN1 or SMAD3) in 15 patients (7.5%) and ≥1 likely pathogenic variant (LPV; in COL3A1, FBN1, FBN2, LOX, MYH11, SMAD3, TGFBR1, or TGFBR2) in 19 individuals (9.6%), together resulting in 17.1% diagnostic yield. Thirteen PV/LPV were novel. Of PV/LPV-negative patients 47 (23.6%) showed ≥1 variant of uncertain significance (VUS). Twenty-five patients had concomitant variants. In-depth evaluation of reported/calculated variant classes resulted in reclassification of 19.8% of variants.ConclusionVariant classification and refinement are essential for shaping mutational spectra of disease genes, thereby improving clinical sensitivity. Obligate stringent multigene analysis is a powerful tool for identifying genetic causes of clinically related CTDs. Nonetheless, the relatively high rate of PV/LPV/VUS-negative patients underscores the existence of yet unknown disease loci and/or oligogenic/polygenic inheritance.     

High-depth next-generation sequencing panel testing in the evaluation of arteriovenous malformations

Arteriovenous malformations (AVMs) are vascular lesions in which an overgrowth of blood vessels of varying sizes develops with one or more direct connections between the arterial and venous circulation. We performed a retrospective review of a cohort of 54 patients with AVMs referred to our clinical genomic laboratory for high-depth next-generation sequencing (NGS) panel of Disorders of Somatic Mosaicism (DoSM). Thirty-seven of 54 patients were female (68.5%). Among the 54 cases, 37 (68.5%) cases had pathogenic and/or likely pathogenic (P/LP) variants identified, two cases (3.7%) had variants of uncertain clinical significance, and the remaining 15 cases (27.8%) had negative results. MAP2K1 variants were found in 12 cases, followed by eight cases with KRAS variants and seven with TEK variants, and the remainder being identified in several other genes on the panel. Among the 37 positive cases, 32 cases had somatic alterations only; the remaining five cases had at least one germline P/LP variant, including four cases with PTEN and one with RASA1. Of note, two cases had the unexpected co-existence of two P/LP variants. In summary, this study illustrated the molecular diagnostic yield (68.5%) of this cohort of patients with a clinical indication of AVMs by our high-depth DoSM NGS panel.     

Molecular basis of carotid body tumor and associated clinical features in Japan identified by genomic,immunohistochemical, and clinical analyses

Carotid body tumor (CBT) is classified as a paraganglioma (PGL). Here, we report the genetic background, protein expression pattern, and clinical findings of 30 Japanese CBT cases. Germline pathogenic or likely pathogenic (P/LP) variants of genes encoding succinate dehydrogenase subunits (SDHs) were detected in 15 of 30 cases (50%). The SDHB variants were the most frequently detected, followed by SDHA and SDHD variants. One case with SDHAF2 variant was bilateral CBT, and other two multiple PGL cases were not detected P/LP variants. The three cases with germline variants that could be tested did not have somatic P/LP variants of the same genes. Immunohistochemical analysis showed negative SDHB signals in CBT tissues in five cases with germline P/LP variants of SDHB, SDHD, or SDHA. In addition, SDHB signals in CBT tissues were negative in four of nine cases without germline P/LP variants of SDHs. These findings suggest the involvement of unidentified molecular mechanisms affecting SDHs.     

Sequential targeted exome sequencing of 1001 patients affected by unexplained limb-girdle weakness

PurposeSeveral hundred genetic muscle diseases have been described, all of which are rare. Their clinical and genetic heterogeneity means that a genetic diagnosis is challenging. We established an international consortium, MYO-SEQ, to aid the work-ups of muscle disease patients and to better understand disease etiology.MethodsExome sequencing was applied to 1001 undiagnosed patients recruited from more than 40 neuromuscular disease referral centers; standardized phenotypic information was collected for each patient. Exomes were examined for variants in 429 genes associated with muscle conditions.ResultsWe identified suspected pathogenic variants in 52% of patients across 87 genes. We detected 401 novel variants, 116 of which were recurrent. Variants in CAPN3, DYSF, ANO5, DMD, RYR1, TTN, COL6A2, and SGCA collectively accounted for over half of the solved cases; while variants in newer disease genes, such as BVES and POGLUT1, were also found. The remaining well-characterized unsolved patients (48%) need further investigation.ConclusionUsing our unique infrastructure, we developed a pathway to expedite muscle disease diagnoses. Our data suggest that exome sequencing should be used for pathogenic variant detection in patients with suspected genetic muscle diseases, focusing first on the most common disease genes described here, and subsequently in rarer and newly characterized disease genes.     

eyeVarP: A computational framework for the identification of pathogenic variants specific to eye disease

PurposeDisease-specific pathogenic variant prediction tools that differentiate pathogenic variants from benign have been improved through disease specificity recently. However, they have not been evaluated on disease-specific pathogenic variants compared with other diseases, which would help to prioritize disease-specific variants from several genes or novel genes. Thus, we hypothesize that features of pathogenic variants alone would provide a better model.MethodsWe developed an eye disease–specific variant prioritization tool (eyeVarP), which applied the random forest algorithm to the data set of pathogenic variants of eye diseases and other diseases. We also developed the VarP tool and generalized pipeline to filter missense and insertion-deletion variants and predict their pathogenicity from exome or genome sequencing data, thus we provide a complete computational procedure.ResultseyeVarP outperformed pan disease–specific tools in identifying eye disease–specific pathogenic variants under the top 10. VarP outperformed 12 pathogenicity prediction tools with an accuracy of 95% in correctly identifying the pathogenicity of missense and insertion-deletion variants. The complete pipeline would help to develop disease-specific tools for other genetic disorders.ConclusioneyeVarP performs better in identifying eye disease–specific pathogenic variants using pathogenic variant features and gene features. Implementing such complete computational procedure would significantly improve the clinical variant interpretation for specific diseases.