Increasing the Yield and Utility of Pediatric Genomic Medicine with Exomiser - PROJECT SUMMARY/ABSTRACT Collectively, rare diseases (RD) are common in the population; approximately 300 million people worldwide have an RD. Approximately 80% of these diseases are caused by gene mutations and up to 75% are present at birth or begin in childhood. In aggregate, RDs have a huge impact on health, particularly that of children, but the diagnosis of these diseases remains challenging: roughly 25% of RD patients must wait between 5 and 30 years for a diagnosis, and about half of the initial diagnoses are wrong. For many affected children, definitive diagnosis comes only after a protracted and frustrating odyssey of visits to different specialists. In recent years, whole genome sequencing (WGS) has entered the mainstream of clinical diagnostics, and new methods— such as long-read genome sequence—are becoming common. The variant information provided by WGS can be extremely useful in diagnosis, but diagnostic rates with WGS remain below 50% in most studies, and new algorithms and software are required to take full advantage of the latest genomic technologies for diagnosis of children with RD. Exomiser, our variant prioritization tool, is widely used to inform RD diagnosis—for example, by NIH’s Undiagnosed Disease Project (NIH-UDP) and Undiagnosed Diseases Network (NIH-UDN), England’s National Health Service (NHS), and the 100,000 Genomes Project, as well as numerous clinics, diagnostic laboratories, research teams, and companies offering decision support tools. Our group also leads the Human Phenotype Ontology (HPO), which has become a foundational global resource for phenotype-driven WES/WGS analysis, both as part of Exomiser and on its own. In this renewal, we will improve Exomiser to enhance clinical diagnostics. We will add new variant prioritizers, including structural and non-coding variants detected by long-read sequencing, and an implementation of the latest ACMG (American College of Medical Genetics and Genomics) variant classification guidelines. We will improve Exomiser’s current machine learning model and enable it to incorporate multiomics data (RNA-seq, proteomics, metabolomics). We will extend Exomiser to enable analysis of non-Mendelian modes of inheritance important in RD such as melded phenotypes and digenic inheritance. New artificial intelligence approaches—including large language models and agentic artificial intelligence—will be applied to improve prioritization and explainability. Cell-type-specific genomic interpretation pipelines will contextualize variants in disease-relevant tissues, while improved multi-source evidence integration will enhance diagnostic confidence. Finally, we will disseminate Exomiser via workshops and tutorials to reach an even broader user group in pediatrics and clinical genetics.
