All articles

The unsolved exome ages well.

A negative exomeSequencing of the protein-coding parts of the genome, about one to two percent of it, where most known disease variants sit. from three years ago is worth more today than when it was reported. What a reanalysisLooking at an unresolved case again when new knowledge arrives, such as a new gene-disease link or a reclassified variant. has to carry to be more than a re-run.

A boy was sequenced at four for a developmental disorder with epilepsy. The trio exome was reported negative in 2023: two variants of uncertain significance in genes with no established link to his phenotype, and nothing to act on. He is seven now. The family has not been back, the referring paediatrician has moved on, and the report sits in the record as a closed question. Nobody has looked at the data since.

Should anyone?

What the field knows

The literature answers that question in one direction. Wright and colleagues (2018), reporting from the Deciphering Developmental Disorders study, re-interpreted genome-wide data from more than a thousand families whose first analysis had been negative or inconclusive. They made new diagnoses in a meaningful share of them, and almost every new diagnosis rested on a gene-disease association published after the original report. Liu and colleagues (2019) described the same pattern from a clinical laboratory in the New England Journal of Medicine: a substantial fraction of exomes that were negative at first report received a diagnosis on reanalysis, and new literature was again the dominant reason. Neither study needed new sequencing.

The mechanism is simple. The sequence does not change, and the knowledge used to read it changes constantly. New gene-disease associations are curated every month; the Clinical Genome Resource framework for gene-disease validity (Strande et al., 2017) exists because the list of genes with enough evidence to report keeps moving. Variants in ClinVar are reclassified as laboratories submit new observations. Annotation improves: transcripts are revised, population frequencies are recalculated on larger cohorts, and pipelines learn to call structural variants and repeat expansions that short-read exomes once missed. On the clinical side, the patient grows and the phenotype becomes more specific. A child with delay at four may have a recognisable epilepsy syndrome at seven.

The American College of Medical Genetics and Genomics responded with a points-to-consider statement on re-evaluation and reanalysis (Deignan et al., 2019). It separates two activities: re-evaluation, in which a reported variant is checked against current evidence, and reanalysis, in which the whole dataset is interpreted again against current knowledge. The statement endorses both, notes that neither has an agreed interval, and leaves the question of who is responsible for triggering them open. Most services have left it open too.

Where practice breaks

Reanalysis fails on logistics before it fails on science. The first problem is the trigger. A reanalysis scheduled by the calendar depends on someone keeping a diary of negative reports, and on that someone still being in post when the date arrives. A reanalysis triggered by a request depends on the family coming back, which the family will only do if someone told them to. In practice most negative reports are never revisited, and the ones that are revisited are the ones whose clinician happened to read the right paper.

The second problem is what gets re-run. A laboratory that re-runs its pipeline on the stored data gets a fresh variant list with current annotation. That is useful. It is also not what the original analysts did. They looked at a list, considered perhaps a dozen candidates against the phenotype, dismissed most of them for reasons that seemed obvious at the time, and kept two as uncertain. None of that reasoning is in the pipeline. A re-run that starts from the list rather than from the reasoning repeats every dismissal, including the ones that a new publication should have overturned.

The third problem is the record. The two variants of uncertain significance from 2023 live in a PDF. The genes they sit in are not on any watch list. When one of those genes acquires a published association with epileptic encephalopathy, the publication reaches the people who read that journal and nobody else. The variant does not know it has become interesting, and neither does the case.

Last, the data. Aligned reads are large, and retention policies vary. A service that wants to reanalyse in five years has to have decided today to keep the files, and to keep the consent that permits looking at them again.

How Tulparus handles it

The workbench treats an open case as a standing question rather than a closed report. Every candidate gene and variant that was considered stays in the case, with the evidence for and against it as it stood and the reason it was set aside. Knowledge arrives as versioned releases: curated gene-disease links, variant classifications, population frequencies and the phenotype ontology. Each release is compared with the open cases it could touch.

For the boy above, the chain is concrete. When a curated release adds a gene-disease link for one of his two uncertain genes, the case is checked again and flagged. When a laboratory reclassifies one of his variants, the case is flagged. When a new population frequency changes how rare a variant is, rarity is measured again against the new figure. Each re-check records the knowledge release and the pipeline version it ran on, so a clinician reading it in 2028 can see what was known when, and so two services looking at the same case see the same evidence.

What the re-check produces is a change for a person to read, with its source: this gene now has a curated association with this phenotype, published here, and the case holds a variant in it that was set aside for this reason. The hypotheses the original team wrote down are still there, so the new evidence lands on a frame rather than on a bare list. A score may move. The workbench does not diagnose, and it does not contact the family. It tells the clinician that the closed question has reopened, and the clinician decides what that means for a child who is now seven.

That is the whole intervention, and it is enough. The yield in the reanalysis literature came from knowledge that already existed and that nobody matched to the cases it could resolve. Matching is work a system can do continuously and a person cannot.

What you can do tomorrow

None of this needs software to start.

  • List the negative exomes your service reported in the last five years, with the genes of every variant left as uncertain. Check those genes against GenCC and ClinGen now. Some have moved.
  • Decide in writing who owns re-evaluation and reanalysis in your service, and what triggers each: a request, an interval, a publication. The ACMG statement is a reasonable template for the conversation.
  • When you dismiss a candidate, record why, in a sentence the next analyst can read without you. A dismissal with a reason can be revisited; one without a reason has to be redone.
  • Ask your laboratory what data it retains, for how long, and under which consent. If the answer is the VCF only, say so in the report, so the clinician who orders a reanalysis knows what is possible.

Sources

  1. Wright CF, McRae JF, Clayton S, et al. Making new genetic diagnoses with old data: iterative reanalysis and reporting from genome-wide data in 1,133 families with developmental disorders. Genetics in Medicine, 2018.
  2. Liu P, Meng L, Normand EA, et al. Reanalysis of clinical exome sequencing data. New England Journal of Medicine, 2019.
  3. Deignan JL, Chung WK, Kearney HM, et al. Points to consider in the reevaluation and reanalysis of genomic test results: a statement of the American College of Medical Genetics and Genomics (ACMG). Genetics in Medicine, 2019.
  4. Strande NT, Riggs ER, Buchanan AH, et al. Evaluating the clinical validity of gene-disease associations: an evidence-based framework developed by the Clinical Genome Resource. American Journal of Human Genetics, 2017.
  5. Richards S, Aziz N, Bale S, et al. Standards and guidelines for the interpretation of sequence variants: a joint consensus recommendation of the ACMG and the Association for Molecular Pathology. Genetics in Medicine, 2015.

Bring one negative exome.See what has changed since.

A demo takes 20 minutes, screen shared, on synthetic cases. Bring the question from one of your own.