The case that stayed open
A family has been through exome or genome sequencing and the report came back without an answer. In many of these cases the answer is not hiding in a gene that is known to cause the patient’s condition. It may sit in a gene that has no established disease link yet, or in a gene whose link is thin. Those genes are the reason a case is worth a second look, and they are also the part that is easiest to get wrong.
Established gene and disease links come with curated evidence: reviewed cases, expert panels, years of reports. A gene with no link has none of that. If both kinds of gene are placed in one ranked list, the list implies they stand on the same ground, and they do not.
Two lists
In the case record, candidate new genes are shown as their own list. They are never merged into the main ranking of established findings, and they are labelled as candidates for expert review. The main list says where the evidence already is. The candidate list says where an expert might want to look next.
Keeping them apart also keeps the main list clean. A clinician who wants only what is established can read only that. A clinician who is out of ordinary options can open the second list on purpose, knowing what it is.
Reasons you can check
A bare name in a list is not worth much. A candidate gene is shown with the reasons it was put forward, and each reason is something a person can go and look at. A reason might be that the gene’s product works with proteins already tied to features this patient has, that it sits in a pathway disrupted in related conditions, or that changing it in a model organism produces something like the patient’s picture. Which reasons appear depends on the gene and on the data that exists for it.
These reasons are put together from curated data. They are not text written by a language model to sound convincing. If a reason is shown, there is a record behind it, and the record can be opened.
What we do not say
A candidate is not a likely cause. A gene can have every reason on that list and still have nothing to do with this patient, because plausibility is cheap and many genes are plausible. We do not give a chance or a hit rate for a candidate, and the list is not a ranking of probabilities.
A person decides what to do with it. That may be a closer look at the variant, a question to a colleague who knows the gene, a functional follow-up, or nothing at all. Whatever they decide is recorded with the case, and the candidate only becomes part of an interpretation when a person signs it as part of one.
Why this is worth being careful about
The pull in this area is toward a long, impressive list. The more candidates a tool shows, the more a team feels it has covered. We would rather show a short list with reasons that hold up when someone reads them, and say plainly that it is a list for experts to review.
How a person might use the list
Picture a case that has been through the established list twice and still has nothing convincing. The team opens the candidate list on purpose. The first entry has a reason that the clinician finds interesting: the gene sits next to a protein that is already linked to a similar picture. They open the sources behind that reason, read two of the underlying records and decide it is worth asking the laboratory to look at the variant in that gene more closely.
The second entry sits just below the first, but its reasons are about general expression in a tissue the patient’s features do not involve. The clinician sets it aside, and the decision to set it aside is recorded with the case, so the next person does not repeat the work.
Neither of those steps is the product reaching a conclusion. Both are people using a list that is honest about what it is.
Review, not discovery
It is tempting to talk about this kind of feature as discovery. We try not to. Finding a gene that has never been linked to a condition takes more than a list: functional work, more families, independent reports. A candidate list is a way to spend a team’s attention a little better, and to leave a record of what was considered. That is a smaller claim and the one we are comfortable making.