EyeG2P: an automated variant filtering approach improves efficiency of diagnostic genomic testing for inherited ophthalmic disorders.

TL;DR: EyeG2P as discussed by the authors is a publically available resource to assist diagnostic filtering of genomic datasets for ophthalmic conditions, utilising the Ensembl Variant Effect Predictor, which enabled a significant increase in precision in comparison to routine testing strategies.

Abstract: PurposeThe widespread adoption of genomic testing for individuals with ophthalmic disorders has increased demand on diagnostic genomic services for these conditions. Moreover, the clinical utility of a molecular diagnosis for individuals with inherited ophthalmic disorders is increasingly placing pressure on the speed and accuracy of genomic testing. MethodsWe created EyeG2P, a publically available resource to assist diagnostic filtering of genomic datasets for ophthalmic conditions, utilising the Ensembl Variant Effect Predictor. We assessed the sensitivity of EyeG2P for 1234 individuals with a broad range of conditions, who had previously received a confirmed molecular diagnosis through routine genomic diagnostic approaches. For a prospective cohort of 83 individuals, we also assessed the precision of EyeG2P in comparision to routine genomic diagnostic approaches. ResultsWe observed that EyeG2P had a 99.5% sensitivity for genomic variants previously identified as a molecular diagnosis for 1234 individuals. EyeG2P enabled a significant increase in precision in comparison to routine testing strategies (p<0.001), with an increased precision in variant analysis of 35% per individual, on average. ConclusionAutomated filtering of genomic variants through EyeG2P can increase the efficiency of diagnostic testing for individuals with a broad range of inherited ophthalmic disorders.

Summary

Introduction

• Inherited ophthalmic disorders are a major cause of blindness in children and working age adults.
• 1,2 Obtaining a genetic diagnosis in affected individuals can inform management, and clinical genomic testing is increasingly being used as a frontline diagnostic tool for these disorders.
• 3-8 Notably, the more widespread availability of gene-directed interventions including gene therapy and preimplantation genetic testing has increased both the value and risk of genomic testing.
• This places substantial demands on the delivery of testing in a timely and accurate manner.
• Here the authors describe and evaluate the diagnostic utility of EyeG2P, a publically available resource for the analysis of genomic variants identified in genes known as a cause of inherited ophthalmic conditions.

Methods

• The G2P web portal (https://www.ebi.ac.uk/gene2phenotype/)15 was used to develop and curate the ophthalmic disorders panel.
• New entries were initiated by selection of a relevant gene symbol from the list of preloaded genes (with their associated Ensembl identifiers).
• These connections were made after inspecting MEDLINE (through the PubMed interface); search terms included the gene name (HGNC) and the disease name (as a minimum).
• A disease mechanism was defined as both an allelic requirement (mode of inheritance, for example biallelic or monoallelic) and a mutation consequence (mode of pathogenicity, for example lossof-function).
• 15 Each locus-genotypemechanism-disease-evidence link was further characterized by assigning to it a set of phenotype terms (i.e. clinical signs and symptoms) from the Human Phenotype Ontology (HPO).

Sequencing and variant identification

• All genomic sequencing datasets were generated in a tertiary healthcare setting (North West Genomic Laboratory Hub, Manchester, UK; ISO 15189:2012; UKAS Medical reference 9865).
• All data collected is part of routine clinical care.
• Analyses to improve genomic diagnostic services for individuals with inherited ophthalmic conditions, as reported in this study, have been approved by the North West Research Ethics Committee (11/NW/0421 and 15/YH/0365).
• Raw sequencing reads were aligned to the GRCh37 reference genome using BWA-mem,18 with single nucleotide variants (SNVs) and indels identified using GATK.19 Larger and more complex indels were identified using Pindel, and copy number variants (CNVs) were identified using DeCON.20 Variants were filtered using quality and read depth thresholds as well as inhouse allele frequencies.
• The zygosity of CNVs were estimated based on their relative read depths.

Routine diagnostics

• Routine genomic analysis was performed utilizing the Congenica platform.
• This process involves filtering variants based on gene/location depending on the gene panel applied, population frequency and predicted molecular consequence.
• A complete list of presets for variant filtering are available in Supplementary Tables 1&2.
• After pre-filtering, variants were analysed by clinically accredited scientists and variants classified in accordance with the 2015 American College of Medical Genetics and Genomics (ACMG) best practice guidelines.

EyeG2P

• Merged VCF files containg SNVs, indels and CNVs were annotated using the G2P plugin for Ensembl Variant Effect Predictor.15,23.
• This plugin requires an input file which lists genes of interest and their allelic requirements; the authors utilized the EyeG2P dataset and an allele frequency cutoff of 0.001 for variants in monoallelic genes and 0.05 for variants in biallelic genes.
• Prospectively collected data from 83 individuals were also used for comparison.
• The authors assessed the sensitivity and the precision of EyeG2P in comparison to results from routine diagnostic analysis.
• All statistical analyses were performed in R and graphics created in R and BioRender.

Results

• Curation of the literature identified 667 genes for inclusion in EyeG2P Between April 2017 and June 2020, the authors interogated the biomedical literature for genes associated with highly penetrant genetic ophthalmic disorders.
• Within the 559 confirmed gene-disease pairs, the associated inheritance patterns were autosomal dominant in 155, autosomal recessive in 341 , X-linked in 31 , and other patterns (including both autosomal dominant and recessive) in 32 instances.
• The authors assessed the capability of EyeG2P to identify molecular diagnoses in 1234 individuals who had previously undergone diagnostic genetic testing at the North West Genomic Laboratory Hub (Manchester, UK).
• The 1267 variants prioritized by EyeG2P were identified in 166 distinct genes and had diverse predicted molecular consequences .
• For 31/33 individuals (94%), the confirmed molecular diagnosis was highlighted by EyeG2P; an average reduction of 7.4 variants for analysis was possible in each individual .

Discussion

• Characterising the genomic basis of inherited ophthalmic conditions has been shown to inform the management of individuals with these conditions.
• The expansion from single gene based testing methodologies to the routine use of large gene panels, exome and genome sequencing approaches requires that robust and accurate informatics filtering strategies are applied to the generated datasets.
• The authors have released these data as a freely available resource that can be dynamically filtered and revised to best aid the users requirements.
• The authors ability to detect disease-causing genomic variants from high-throughput sequencing datasets has expanded in recent years to include complex structural variants,26,27 exonic deletions and duplications,28,29 deeply intronic variants causing aberrant splicing,30-33 variants in regulatory regions34-36 and complex alleles comprised of combinations of genomic variants common in the general population.
• Moreover, EyeG2P can identify diverse genomic variants across the spectrum of genetically and clinically heterogeneous ophthalmic genetic conditions.

Figures

Figure 2. Direct comparison between EyeG2P and routine testing approaches identifies increased precision and efficiency of EyeG2P as a first-tier analysis approach. (a) the number of variants requiring analysis by clinical scientists for 83 individuals receiving genetic testing for inherited Ophthalmic disorders. (b) summary of the molecular consequences of disease-causing variants identified to underpin confirmed molecular diagnoses in 31 individuals. (c) the precision (PPV, positive predictive value) of different testing approaches for 31 individuals receiving a diagnosis through both approaches.

Figure 1. Molecular findings for 1228 individuals with a confirmed molecular diagnosis for inherited Opthalmic disorders. (a) EyeG2P is a plugin for the Ensembl Variant Effector enabling automated variant filtering and selection of variants in a disease-causing state. The specific requirements of variants to be retained can be set by the user and through the inclusion of predefined list of genomic variants, including pathogenic variants in ClinVar, variants predicted to impact splicing, and complex alleles comprised of variants above the defined variant frequency threshold. (b) The predicted molecular consequences of 1267 variants identified as a cause of disease in 1228 individuals, demonstrating the wide range of variant consequences that can be prioritized by EyeG2P. (c) The number of variants identified as a cause of disease in 166 genes by their proven zygosity. Hemizygous variants are included in the Homozygous display.

Full text

EyeG2P: an automated variant filtering approach improves efficiency
of diagnostic genomic testing for inherited ophthalmic disorders.
Eva Lenassi
1,2,3
, Ana Carvalho
4,5
, Anja Thormann
6
, Tracy Fletcher
2
, Claire Hardcastle
2
, Sarah E Hunt
6
,
Panagiotis I Sergouniotis
1,2,3
, Michel Michaelides
6,7
, Andrew R Webster
6,7
, Cunningham F
6
, Simon
Ramsden
2
, David R FitzPatrick
4
, Graeme CM Black
1,2
, Jamie M Ellingford
1,2
1. Division of Evolution and Genomic Sciences, School of Biological Sciences, Faculty of Biology, Medicine
and Health, University of Manchester, Manchester, United Kingdom.
2. Manchester Centre for Genomic Medicine, St Mary's Hospital, Manchester University NHS Foundation
Trust, Manchester, United Kingdom.
3. Manchester Royal Eye Hospital, Manchester University NHS Foundation Trust, Manchester, United
Kingdom.
4. MRC Human Genetics Unit, MRC Institute of Genetics and Cancer, University of Edinburgh, Edinburgh,
UK.
5. Medical Genetic Unit, Pediatric Hospital, Coimbra Hospital and Universitary Centre (CHUC), Coimbra,
Portugal
6. European Molecular Biology Laboratory, European Bioinformatics Institute, Wellcome Genome
Campus, Hinxton, CB20 1SD, Cambridge, UK.
7. UCL Institute of Ophthalmology, University College London, London, UK.
8. Department of Ophthalmology, Moorfields Eye Hospital NHS Foundation Trust, London, UK.
Correspondence
graeme.black@manchester.ac.uk
jamie.ellingford@manchester.ac.uk
. CC-BY 4.0 International licenseIt is made available under a
is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)
The copyright holder for this preprint this version posted July 25, 2021. ; https://doi.org/10.1101/2021.07.23.21261017doi: medRxiv preprint
NOTE: This preprint reports new research that has not been certified by peer review and should not be used to guide clinical practice.

Abstract(
Purpose: The widespread adoption of genomic testing for individuals with ophthalmic disorders has
increased demand on diagnostic genomic services for these conditions. Moreover, the clinical utility of a
molecular diagnosis for individuals with inherited ophthalmic disorders is increasingly placing pressure on
the speed and accuracy of genomic testing.
Methods: We created EyeG2P, a publically available resource to assist diagnostic filtering of genomic
datasets for ophthalmic conditions, utilising the Ensembl Variant Effect Predictor. We assessed the
sensitivity of EyeG2P for 1234 individuals with a broad range of conditions, who had previously received
a confirmed molecular diagnosis through routine genomic diagnostic approaches. For a prospective
cohort of 83 individuals, we also assessed the precision of EyeG2P in comparision to routine genomic
diagnostic approaches.
Results: We observed that EyeG2P had a 99.5% sensitivity for genomic variants previously identified as a
molecular diagnosis for 1234 individuals. EyeG2P enabled a significant increase in precision in comparison
to routine testing strategies (p<0.001), with an increased precision in variant analysis of 35% per
individual, on average.
Conclusion: Automated filtering of genomic variants through EyeG2P can increase the efficiency of
diagnostic testing for individuals with a broad range of inherited ophthalmic disorders.(
(
. CC-BY 4.0 International licenseIt is made available under a
is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)
The copyright holder for this preprint this version posted July 25, 2021. ; https://doi.org/10.1101/2021.07.23.21261017doi: medRxiv preprint

Introduction(
Inherited ophthalmic disorders are a major cause of blindness in children and working age adults.
1,2
Obtaining a genetic diagnosis in affected individuals can inform management, and clinical genomic testing
is increasingly being used as a frontline diagnostic tool for these disorders.
3-8
Notably, the more
widespread availability of gene-directed interventions including gene therapy and preimplantation
genetic testing has increased both the value and risk of genomic testing.
9-14
This places substantial
demands on the delivery of testing in a timely and accurate manner.
Here we describe and evaluate the diagnostic utility of EyeG2P, a publically available resource for the
analysis of genomic variants identified in genes known as a cause of inherited ophthalmic conditions. We
curated disease-causing genes through robust and transparent standards and assessed the sensitivity and
precision of EyeG2P in a cohort of individuals receiving diagnostic testing for a range of ophthalmic
conditions. EyeG2P uses logical filtering of identified genomic variants in line with their predicted
molecular consequence, population frequency and prior knowledge of disease mechanisms and
inheritance patterns. Overall we found that utilizing EyeG2P as a first-tier analysis strategy reduces the
number of variants requiring analysis by clinically accredited scientists and increases the precision and
efficiency of diagnostic testing for ophthalmic disorders.
. CC-BY 4.0 International licenseIt is made available under a
is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)
The copyright holder for this preprint this version posted July 25, 2021. ; https://doi.org/10.1101/2021.07.23.21261017doi: medRxiv preprint

Methods (
Curation(of ( known(disease(genes(
The G2P web portal (https://www.ebi.ac.uk/gene2phenotype/)
15
was used to develop and curate the
ophthalmic disorders panel. New entries were initiated by selection of a relevant gene symbol from the
list of preloaded genes (with their associated Ensembl identifiers). For each entry, a gene or locus was
linked, via a disease mechanism, to a disease. These connections were made after inspecting MEDLINE
(through the PubMed interface); search terms included the gene name (HGNC) and the disease name (as
a minimum). A disease mechanism was defined as both an allelic requirement (mode of inheritance, for
example biallelic or monoallelic) and a mutation consequence (mode of pathogenicity, for example loss-
of-function). A confidence attribute—confirmed, probable or possible—was also assigned to indicate how
likely it is that the gene is implicated in the cause of disease; the rules used to assign confidence, allelic
requirement and mutation consequence to entries are defined and available.
15
Each locus-genotype-
mechanism-disease-evidence link was further characterized by assigning to it a set of phenotype terms
(i.e. clinical signs and symptoms) from the Human Phenotype Ontology (HPO).
16
The details of the edits
and the identifiers of the relevant publications (that provide evidence for that specific gene-disease
thread) were stored and are available through the G2P web portal.
Sequencing(and(variant(identification(
All genomic sequencing datasets were generated in a tertiary healthcare setting (North West Genomic
Laboratory Hub, Manchester, UK; ISO 15189:2012; UKAS Medical reference 9865). Individuals provided
written consent for genomic analysis and all investigations were conducted in accordance to the tenets
of the Declaration of Helsinki. All data collected is part of routine clinical care. Analyses to improve
genomic diagnostic services for individuals with inherited ophthalmic conditions, as reported in this study,
have been approved by the North West Research Ethics Committee (11/NW/0421 and 15/YH/0365). No
individual patient data is reported in this manuscript.
Routine diagnostic gene panel testing was performed as previously described.
3,5,17
Briefly, DNA samples
were processed using Agilent SureSelect (Agilent Technologies, Santa, Clara, CA) target enrichment kits
designed to capture selected intronic regions and all protein-coding exons +/-50 base pairs of flanking
intronic sequences of selected panels of known disease-genes. The decision on which panel to use was
made by the referring clinician (either a consultant ophthalmologist or a consultant clinical geneticist with
an interest in ophthalmic genetics).
. CC-BY 4.0 International licenseIt is made available under a
is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)
The copyright holder for this preprint this version posted July 25, 2021. ; https://doi.org/10.1101/2021.07.23.21261017doi: medRxiv preprint

Sequencing was performed using Illumina HiSeq and NextSeq platforms. Raw sequencing reads were
aligned to the GRCh37 reference genome using BWA-mem,
18
with single nucleotide variants (SNVs) and
indels identified using GATK.
19
Larger and more complex indels were identified using Pindel, and copy
number variants (CNVs) were identified using DeCON.
20
Variants were filtered using quality and read
depth thresholds as well as inhouse allele frequencies. The zygosity of CNVs were estimated based on
their relative read depths. Regions that are highly polymorphic and/or difficult to survey through short-
read high-throughput techniques are masked from initial analysis, specifically RP1L1 exon 4, USH1C exon
18 and RPGRorf15.
Variant(analysis(
Routine diagnostics
Routine genomic analysis was performed utilizing the Congenica platform. This process involves filtering
variants based on gene/location depending on the gene panel applied, population frequency and
predicted molecular consequence. A complete list of presets for variant filtering are available in
Supplementary Tables 1&2. After pre-filtering, variants were analysed by clinically accredited scientists
and variants classified in accordance with the 2015 American College of Medical Genetics and Genomics
(ACMG) best practice guidelines.
21,22
EyeG2P
Merged VCF files containg SNVs, indels and CNVs were annotated using the G2P plugin for Ensembl
Variant Effect Predictor.
15,23
This plugin requires an input file which lists genes of interest and their allelic
requirements; we utilized the EyeG2P dataset and an allele frequency cutoff of 0.001 for variants in
monoallelic genes and 0.05 for variants in biallelic genes. An additional list was used as input including all
ClinVar pathogenic or likely pathogenic variants, all variants predicted to have a significant impact on
splicing by SpliceAI,
24
and a selection hypomorphic alleles that are known to be pathogenic but exceed
the variant frequency thresholds specified.
Comparisons between routine diagnostic analysis and EyeG2P
Results from EyeG2P were restrospectively compared to clinically reported variants identified from
routine diagnostic analysis in 1234 individuals with genomic opthalmic conditions. All these study
participants had a confirmed (or a provisional) molecular diagnosis and carried pathogenic (ACMG class
5), likely pathogenic (ACMG class 4) or variants of uncertain significance (ACMG class 3); these changes
. CC-BY 4.0 International licenseIt is made available under a
is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)
The copyright holder for this preprint this version posted July 25, 2021. ; https://doi.org/10.1101/2021.07.23.21261017doi: medRxiv preprint

Frequently asked questions

Q1. What are the contributions in "Eyeg2p: an automated variant filtering approach improves efficiency of diagnostic genomic testing for inherited ophthalmic disorders" ?

The authors created EyeG2P, a publically available resource to assist diagnostic filtering of genomic datasets for ophthalmic conditions, utilising the Ensembl Variant Effect Predictor. It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. 

Q2. What plugin was used to annotate variants?

21,22Merged VCF files containg SNVs, indels and CNVs were annotated using the G2P plugin for Ensembl Variant Effect Predictor.15,23 

Q3. What is the definition of a disease mechanism?

A disease mechanism was defined as both an allelic requirement (mode of inheritance, for example biallelic or monoallelic) and a mutation consequence (mode of pathogenicity, for example lossof-function). 

Q4. What is the importance of a robust and accurate informatics filtering strategy?

The expansion from single gene based testing methodologies to the routine use of large gene panels, exome and genome sequencing approaches requires that robust and accurate informatics filtering strategies are applied to the generated datasets. 

Q5. How many variants were identified in the study?

Disease-causing variants were identified in 24 distinct genes; 10 cases had an autosomal dominant, 3 an X-linked and 18 an autosomal recessive disorder. 

Q6. How many cases of ocular diseases were identified without a diagnosis?

In 10/52 cases without a confirmed diagnosis, variants of uncertain significance were identified in a disease-causing state through EyeG2P analysis, and no additional pathogenic variants were detected after routine analysis. 

Q7. What is the genetic basis of ophthalmic conditions?

The genetic basis of ophthalmic conditions such as congenital cataract, inherited retinal disorders and optic neuropathies is diverse and includes genes encoded on autosomal, sex and mitochondrial chromosomes. 

Q8. What is the purpose of this paper?

In conclusion, the authors demonstrate that EyeG2P can be effectively integrated with clinical diagnostic testing for inherited ophthalmic conditions to increase the efficiency of variant analysis. 

Q9. What was the PPV of the EyeG2P analysis?

Following curation of over 1000 biomedical publications the authors identified 667 relevant genes and determined the associated modes of inheritance, mechanisms of disease causation and phenotypic features. 

Q10. What are the sources of funding for EyeG2P?

The authors acknowledge funding from the Wellcome Trust Transforming Genomic Medicine Initiative (WT200990/Z/16/Z), the European Molecular Biology Laboratory and the Manchester NIHR Biomedical Research Centre (IS-BRC-1215-20007). 

Q11. What is the common type of variant that is masked from the initial analysis?

Regions that are highly polymorphic and/or difficult to survey through shortread high-throughput techniques are masked from initial analysis, specifically RP1L1 exon 4, USH1C exon 18 and RPGRorf15. 

Q12. What is the purpose of the paper?

The authors propose the application of EyeG2P as a firsttier analysis strategy for the diagnosis of inherited opthalmic conditions from high-throughput genomic datasets. 

Q13. What is the plugin for Ensembl Variant Effect Predictor?

This plugin requires an input file which lists genes of interest and their allelic requirements; the authors utilized the EyeG2P dataset and an allele frequency cutoff of 0.001 for variants in monoallelic genes and 0.05 for variants in biallelic genes. 

Q14. What is the difference between EyeG2P and the standard diagnostic testing methods?

The authors show that EyeG2P increases the precision and efficiency of genomic testing for inherited ophthalmic conditions over routine approaches for variant analysis, at little cost to overall diagnostic rates.