ABSTRACT We propose to study the performance of GenomeDiver in helping to make diagnoses from whole genome sequencing (WGS) in children with neurological, cardiological and immunological diseases. This project will focus on ~400 children left without a diagnosis following the NYCKidSeq study, part of the NHGRI CSER consortium focused on implementing diagnostic WGS in diverse populations. GenomeDiver was developed as part of the NYCKidSeq study, with the goal of improving the ability of the clinical geneticist to provide phenotypic information as part of the diagnostic process. GenomeDiver is a digital medicine application that uses as its input the patient’s genomic sequence information (as a variant call format (VCF) file) and the phenotypic information made available to the diagnostic laboratory. The software embeds Exomiser to prioritize variants, allowing the extraction of Human Phenotype Ontology (HPO) terms that characterize and distinguish the highest-ranked variants. These HPO terms are presented within the GenomeDiver interface to the clinical geneticist who categorizes them as Present, Absent or Uncertain in the patient. The enhanced phenotypic information is then used to re-run Exomiser, which then presents the gene names and associated diseases to the clinician, who can flag any of interest before returning all the information to the diagnostic laboratory to augment the information they can use diagnostically. In this project, we will use GenomeDiver on diverse Bronx patients from the Montefiore Health System who have participated in our NYCKidSeq project. Approximately 74% have been left without a diagnosis, a common problem in diagnostic sequencing even when all exons or the entire non-repetitive genome is sequenced. We note that both the patient’s phenotype and the discovery of new pathogenic variants evolves over time, and that re-analysis should be expected to permit new diagnoses to be made in some patients. We will therefore divide ~400 patients into two groups, one of which will have a GenomeDiver intervention added to the standard of care. This pilot study is designed primarily to get feedback from clinician users about the design and utility of the software, allowing its further refinement. We will also compare diagnostic yield in the two groups, generating an estimate of the 95% confidence interval that will allow us to design a follow up, expanded multicenter trial of GenomeDiver. Our overall goal is to understand how we can implement a provider-facing software app in clinical care of patients with genetic disorders to improve diagnostic yield of WGS. Our process of prompting the clinician to look for specific phenotypic features based on genomic information is unusual, and something we describe as ‘reverse phenotyping’. Part of our motivation is to demonstrate to clinicians that reverse phenotyping is a practical and valuable component of the diagnostic process, and that a tool such as GenomeDiver can be part of the decision support in the care of complex genetic disorders.

NARRATIVE In this project, we will test how the GenomeDiver software performs in a clinical environment, using it in the follow up of children sequenced as part of the NYCKidSeq study on whom a diagnosis was not made from their initial testing. GenomeDiver is designed to improve the ability of the clinician caring for the patient to provide a greater amount and quality of phenotypic information to help with the diagnostic process. Our primary outcome will be the feedback collected from the clinical users as a guide to further development of the software, with a secondary outcome to analyze any improvement in the rate of diagnosis using GenomeDiver, as the foundation for a follow up, expanded clinical trial to test its use in genomic diagnostics.