A boy of five with a severe developmental disorder, microcephaly and a distinctive electroencephalogram has a trio exome. The known-gene analysis is negative. One de novo variant remains: a missense change in a gene with no disease association, highly intolerant to variation in population data, expressed in the developing brain, and encoding a protein in the same complex as two established neurodevelopmental genes.
The laboratory cannot report it as pathogenic; the framework does not allow that for a gene with no established disease. It cannot dismiss it either. The variant goes into a sentence at the end of the report that no one acts on, and the boy's case is closed as negative with its best lead inside it.
What the field knows
The variant interpretation framework (Richards et al., 2015) classifies variants in genes whose relationship to disease is already established. It says so explicitly, and the ClinGen gene-disease validity framework (Strande et al., 2017) is the separate process by which a gene earns that status, through accumulated genetic and experimental evidence scored from limited to definitive. A variant in a gene with no validity classification sits outside the first framework, and a single patient cannot by themselves move the gene through the second.
What a single patient can do is contribute. MacArthur and colleagues (2014) set out the standards for implicating a sequence variant in disease, and the central point was that a candidate in a novel gene is a hypothesis that needs independent evidence: a second unrelated patient with a matching phenotype and a comparable variant, functional data, or a model. Matchmaker Exchange (Philippakis et al., 2015) was built for the first of these, connecting candidate genes across laboratories so that the second patient can be found, and a large share of new disease genes in the years since have come through matches of this kind.
The signals that make a novel candidate worth pursuing are known and quantifiable. Constraint against missense or loss-of-function variation in population data (Lek et al., 2016) identifies genes in which a damaging variant is unlikely to be tolerated. De novo status in a sporadic severe disorder, expression in the relevant tissue at the relevant time, membership of a complex or pathway with established disease genes, and a model organism phenotype that resembles the patient's each raise the prior. None of them is a diagnosis. Together they are a reason to look for the second patient.
Where practice breaks
The novel candidate falls between two processes. The clinical laboratory's job is classification under the framework, and the framework has nothing to say. The research laboratory's job is discovery, and the research laboratory does not have the case. The variant is mentioned in the report as a research finding and the mention is the end of its clinical life.
The standard for pursuing a candidate is not written down, so it is applied inconsistently. One scientist submits every de novo in a constrained gene to a matchmaking network; another submits none, on the grounds that it is not clinical work; a third submits when the family asks. Which patients get a second look depends on who happened to review them.
Time works against the lead. The gene that was novel at the first analysis may be established two years later by other laboratories' patients, and the case that carried the first observation is not reopened because nobody kept the candidate as an open question with a gene to watch. The patient who could have been the second case in a publication is still filed as negative.
Where a candidate is submitted, the submission is detached from the case. It lives in a matchmaking portal under a scientist's account. If a match comes two years later, the email goes to someone who may have left, and the case it refers to has to be found by hand. If the scientist moves, the submission is orphaned.
The risk the clinical laboratory is right to fear is over-reading. A novel candidate presented to a family with the language of a diagnosis is harmful: it ends the search, it can drive decisions, and it may be wrong. The answer to that risk has been to say as little as possible, which protects the laboratory and loses the lead. The alternative is to say exactly what the candidate is, with the standard it has and has not met, and to keep working on it.
How Tulparus handles it
The workbench keeps two kinds of hypothesis apart. A hypothesis on an established gene is assessed under the variant framework and can reach a reportable classification. A hypothesis on a novel gene is assessed under a separate, explicit standard: the evidence that would be needed for the gene to reach limited or moderate validity, and how much of it this case contributes. The page says which kind it is, and the two are never presented side by side as if they were the same claim.
The novel-gene standard is written into the case before the candidate is pursued. It lists the signals the candidate has, constraint, de novo status, expression, pathway, model organism, each with its source and version, and the evidence still missing, which is almost always a second patient and often a functional result. A clinician reading the hypothesis sees a candidate with a named gap, and a reader of the report sees a research lead described as one, with the standard stated in words.
The second patient is sought as a next step on the case rather than as a side activity. Where the laboratory has opted in, the workbench prepares a matchmaking submission from the case with the minimum it needs, records that it was made and when, and routes any match back to the case rather than to an individual's inbox. If the scientist leaves, the case still knows the submission exists.
For the boy above, the case would stay open on a novel-gene hypothesis with its standard and its gap, the submission would be recorded on the case, and when a second laboratory submitted a matching patient, the match would arrive on his case with the comparison laid out, for a person to assess.
What you can do tomorrow
Write the standard before the next case needs it.
- Agree, in writing, which novel candidates the laboratory will pursue: for example, a de novo variant in a highly constrained gene in a severe sporadic disorder with a plausible tissue and pathway link. Apply it to every case.
- Keep the matchmaking submission in the case record, with the date and the identifier, and name the person who receives a match if the submitter leaves.
- Report a novel candidate as a research lead in plain words, with what it has and has not met, so that the family and the clinician know what it is and the search does not stop.
- Review open novel-gene hypotheses when gene-disease validity resources release, since the gene may have been established by someone else's patients since you last looked.
Sources
- Richards S, Aziz N, Bale S, et al. Standards and guidelines for the interpretation of sequence variants: a joint consensus recommendation of the American College of Medical Genetics and Genomics and the Association for Molecular Pathology. Genetics in Medicine, 2015.
- 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.
- MacArthur DG, Manolio TA, Dimmock DP, et al. Guidelines for investigating causality of sequence variants in human disease. Nature, 2014.
- Philippakis AA, Azzariti DR, Beltran S, et al. The Matchmaker Exchange: a platform for rare disease gene discovery. Human Mutation, 2015.
- Lek M, Karczewski KJ, Minikel EV, et al. Analysis of protein-coding genetic variation in 60,706 humans. Nature, 2016.