Most patients with a suspected genetic condition still move through testing one layer at a time. A panel checks a set of genes, a separate assay looks for a repeat expansion, another studies structural changes. Each step answers part of the question and leaves the rest of the genome out of view. Diagnostic journeys stretch across years, and some end without an answer at all.
In the most recent episode of The Genetics Podcast, Patrick Short spoke with Dr. Danny Miller, Assistant Professor at the University of Washington and an attending physician at Seattle Children's Hospital. Their conversation worked through why that fragmented path persists and what a single, complete genome could do in its place.
Long-read sequencing reads long, continuous stretches of DNA in a single run. That lets one assay capture SNPs, indels, structural variants, repeat expansions, methylation, and variant phasing together. Against short-read sequencing, it finds at least roughly 50% more structural variants, the class of change that narrower tests most often miss.
For Danny, that completeness changes what should be offered first. As he put it: "I've gotten to the point where I think every individual who has a suspected Mendelian condition should be offered the best, most comprehensive test that we have available today, and that's long-reads."
The clinical payoff shows up in cases that other methods leave open. One cohort included eight children from seven families with incomplete or no molecular diagnosis. Long-read sequencing resolved a roughly 2,600 base-pair SVA-E transposable-element insertion that created a novel splice site, explaining the Canavan disease that earlier testing had missed.
Three years ago, cost was a fair objection to running long-read first. That objection has closed. A sub-$500 human genome from PacBio now puts long-read on par with short-read pricing, and both PacBio and Nanopore continue to push the cost down.
What remains is the work of reading the data and getting it reimbursed. Danny framed the current constraint around interpretation and reporting: the effort now sits in making sense of the variants a complete test returns, and in the reimbursement incentives that do not yet reward that added completeness. That, in his view, is the hurdle to clear before long-read reaches patients in the clinic.
Reading a structural variant means comparing it against a population, and long-read reference data has been scarce. Danny's lab is building a public long-read dataset from the 1000 Genomes cohort to fill that gap, and the filtering power is already clear.
He described the effect in concrete terms: "If you just take five hundred individuals from 1000 Genomes, you can filter out more than ninety-five percent of the structural variants you find in that person." For comparison, GnomAD, drawn from roughly 120,000 haplotypes across about 60,000 individuals, filters out about 66% of the structural variants in one person. Around 1,000 individuals from the 1000 Genomes cohort have been sequenced with long-read so far. The goal is to reach all roughly 3,200, and NIH support is secured to continue.
Long-read sequencing also captures methylation, and Danny argued that the resulting epi-signatures belong in the open. Today many signatures sit inside a single company that does not share the underlying data or methods. That makes results hard to replicate and harder to build on for the rare disease community.
Made public, methylation signatures can support diagnosis and, looking ahead, measure whether a therapy is working. Danny sees them as an objective readout for gene therapies and antisense oligonucleotides. They offer a way to confirm biological effect rather than infer it from symptoms.
Danny's longer view is a genome that sits quietly in the record and gets consulted like any other result. The clearest test of that idea is the NICU. He described admitting a newborn at 8am and having the genome back by noon. An antisense oligonucleotide candidate could be identified shortly after, with no new scientific breakthrough required.
He was precise about where the remaining work sits: "There is no technical limitation to getting information, to having a child's genome sequenced and having actionable information within hours of collecting sample. It's simply an operational problem at this point." Closing that gap depends on interpretation pipelines, reporting standards, and training the clinicians who will use the data.
The through-line of the conversation is that the science needed for first-line long-read sequencing largely exists. Progress now depends on interpretation, reimbursement, open reference data, and clinician training rather than a new instrument. That reframes the diagnostic odyssey as a system to redesign. It also puts the tools within reach of the teams willing to do the operational work.
Listen to the full episode below.