Determining the characteristics of genetic disorders that predict inclusion in newborn genomic sequencing programs.

Over 30 international research studies and commercial laboratories are exploring the use of genomic sequencing to screen apparently healthy newborns for genetic disorders. These programs have individualized processes for determining which genes and genetic disorders are queried and reported in newborns. We compared lists of genes from 26 research and commercial newborn screening programs and found substantial heterogeneity among the genes included. A total of 1,750 genes were included in at least one newborn genome sequencing program, but only 74 genes were included on >80% of gene lists, 16 of which are not associated with conditions on the Recommended Uniform Screening Panel. We used a linear regression model to explore factors related to the inclusion of individual genes across programs, finding that a high evidence base as well as treatment efficacy were two of the most important factors for inclusion. We applied a machine learning model to predict how suitable a gene is for newborn sequencing. As knowledge about and treatments for genetic disorders expand, this model provides a dynamic tool to reassess genes for newborn screening implementation. This study highlights the complex landscape of gene list curation among genomic newborn screening programs and proposes an empirical path forward for determining the genes and disorders of highest priority for newborn screening programs.

Evidence review and considerations for use of first line genome sequencing to diagnose rare genetic disorders.

Early use of genome sequencing (GS) in the diagnostic odyssey can reduce suffering and improve care, but questions remain about which patient populations are most amenable to GS as a first-line diagnostic test. To address this, the Medical Genome Initiative conducted a literature review to identify appropriate clinical indications for GS. Studies published from January 2011 to August 2022 that reported on the diagnostic yield (DY) or clinical utility of GS were included. An exploratory meta-analysis using a random effects model evaluated DY based on cohort size and diagnosed cases per cohort. Seventy-one studies met inclusion criteria, comprising over 13,000 patients who received GS in one of the following settings: hospitalized pediatric patients, pediatric outpatients, adult outpatients, or mixed. GS was the first-line test in 38% (27/71). The unweighted mean DY of first-line GS was 45% (12-73%), 33% (6-86%) in cohorts with prior genetic testing, and 33% (9-60%) in exome-negative cohorts. Clinical utility was reported in 81% of first-line GS studies in hospitalized pediatric patients. Changes in management varied by cohort and underlying molecular diagnosis (24-100%). To develop evidence-informed points to consider, the quality of all 71 studies was assessed using modified American College of Radiology (ACR) criteria, with five core points to consider developed, including recommendations for use of GS in the N/PICU, in lieu of sequential testing and when disorders with substantial allelic heterogeneity are suspected. Future large and controlled studies in the pediatric and adult populations may support further refinement of these recommendations.

A framework for the evaluation and reporting of incidental findings in clinical genomic testing.

Currently, there are no widely accepted recommendations in the genomics field guiding the return of incidental findings (IFs), defined here as unexpected results that are unrelated to the indication for testing. Consequently, reporting policies for IFs among laboratories offering genomic testing are variable and may lack transparency. Herein we describe a framework developed to guide the evaluation and return of IFs encountered in probands undergoing clinical genome sequencing (cGS). The framework prioritizes clinical significance and actionability of IFs and follows a stepwise approach with stopping points at which IFs may be recommended for return or not. Over 18 months, implementation of the framework in a clinical laboratory facilitated the return of actionable IFs in 37 of 720 (5.1%) individuals referred for cGS, which is reduced to 3.1% if glucose-6-phosphate dehydrogenase (G6PD) deficiency is excluded. This framework can serve as a model to standardize reporting of IFs identified during genomic testing.

Development of a comprehensive cardiovascular disease genetic risk assessment test.

Background: Despite monogenic and polygenic contributions to cardiovascular disease (CVD), genetic testing is not widely adopted, and current tests are limited by the breadth of surveyed conditions and interpretation burden. Methods: We developed a comprehensive clinical genome CVD test with semi-automated interpretation. Monogenic conditions and risk alleles were selected based on the strength of disease association and evidence for increased disease risk, respectively. Non-CVD secondary findings genes, pharmacogenomic (PGx) variants and CVD polygenic risk scores (PRS) were assessed for inclusion. Test performance was modeled using 2,594 genomes from the 1000 Genomes Project, and further investigated in 20 previously tested individuals. Results: The CVD genome test is composed of a panel of 215 CVD gene-disease pairs, 35 non-CVD secondary findings genes, 4 risk alleles or genotypes, 10 PGx genes and a PRS for coronary artery disease. Modeling of test performance using samples from the 1000 Genomes Project revealed ~6% of individuals with a monogenic finding in a CVD-associated gene, 6% with a risk allele finding, ~1% with a non-CVD secondary finding, and 93% with CVD-associated PGx variants. Assessment of blinded clinical samples showed complete concordance with prior testing. An average of 4 variants were reviewed per case, with interpretation and reporting time ranging from 9-96 min. Conclusions: A genome sequencing based CVD genetic risk assessment can provide comprehensive genetic disease and genetic risk information to patients with CVD. The semi-automated and limited interpretation burden suggest that this testing approach could be scaled to support population-level initiatives.