Wave. Earlier today, I was invited to speak at the Kids First Spring Public Webinar, part of the Gabriella Miller Kids First Pediatric Research Program. The theme was “Ride the Wave of Discovery,” fitting for a program built around shared pediatric data and collaboration. In my talk, I focused on epilepsy genetics and on a question that now sits in front of pediatric genomics: what can we learn when many children already have genomes? My answer was that we need to move from genomes to phenomes to time.
Figure. The Kids First Neuroscience Cohort. For this post, I am breaking one of my own visual rules and using one of my webinar slides rather than converting the figure fully into Channelopathist Blue. The figure shows a so-called beachflag plot, generated by merging data from two Kids First X01 cohorts. The x-axis represents age in years, and the y-axis shows individuals ordered by their first encounter in the electronic medical record. Each horizontal line represents one patient’s available clinical record over time, and each dot represents a single patient encounter. Pediatric brain tumor data are shown in light blue, while epilepsy data are shown in purple. Together, the figure gives a visual sense of what longitudinal clinical data look like before they become analysis-ready: thousands of individual clinical traces, irregularly spaced across childhood, forming the raw material for phenotype science.
Kids First. Kids First was inspired by Gabriella Miller, who was diagnosed at age nine with a rare and aggressive brain tumor and used her illness to advocate for children with serious diseases. After her death, the Gabriella Miller Kids First Research Act was named in her honor and signed into law in 2014 after what was described as a rare bipartisan effort to support pediatric research. The program now supports data-driven research across childhood cancer, structural birth defects, and other pediatric conditions. It represents an infrastructure for pediatric data generation, data sharing, harmonized clinical information, and cloud-based discovery. In our webinar, one number stood out: more than 30,000 genomes from affected children and families are now part of the Kids First data ecosystem.
The question. What can we learn when everybody has a genome? For many years, the main answer in pediatrics was gene discovery. We sequenced families to identify genes in which rare variants cause severe pediatric disease, including rare childhood epilepsies and neurodevelopmental disorders. This work gave names to previously unexplained conditions and created part of the foundation for precision medicine. Once genomes are available at scale, however, the bottleneck starts to shift. The next challenge is to understand what a variant means across development, clinical course, and treatment.
The bottleneck. In my presentation, I talked about the “phenotypic bottleneck.” Genomic data are increasingly standardized and computable. Phenotypes are different. They are often buried in clinical notes, shaped by the setting in which care is delivered, and scattered across years of follow-up. One individual may be described as having developmental delay, another as having delayed speech, and another as being nonverbal. These descriptions may point to the same underlying feature, but without a shared language, they remain difficult to compare. This is where phenotype science comes in.
Phenotype science. Frameworks such as the Human Phenotype Ontology (HPO), which we have worked with extensively, help translate clinical observations into a computable language. Tools using this harmonized language allow us to compare patterns across individuals. In epilepsy genetics, this matters because the field has already moved through an intense period of gene discovery, while we are still learning the core features and boundaries of many rare conditions that were discovered through sequencing. In my talk, I brought up our discovery of AP2M1 again, one of the first epilepsy genes found through systematic analysis of phenotypic similarity across individuals with existing genomic data.
Time. In addition to harmonizing phenotypes, there is another layer. A genome is usually measured once, but disease and natural history happen over time. For a child with a rare epilepsy or neurodevelopmental disorder, the meaningful questions often sit within the longitudinal disease trajectory, rather than within a one-time assessment of clinical features. I therefore highlighted some of our work with longitudinal clinical data to reconstruct seizure burden and medication histories. I also discussed our 2024 work by Galer et al., which showed detectable phenotypic signatures in genes such as SCN1A up to six months before genetic testing is performed.
What you need to know. The genome is really only the beginning. The next wave of pediatric discovery will come from linking genomes to phenomes, and phenomes to time. Kids First shows what this future may look like: large-scale genomic data connected to clinical information, shared across the research community, and used to ask questions that no single dataset could answer alone.
