Paralogs. Every week in our variant review meetings, we encounter a familiar issue: understanding a missense variant of uncertain significance. Unless it matches a known disease-associated variant or is found to be de novo, our confidence often stalls. But what happens if we stopped looking at genes in isolation? In a recent publication, we had the opportunity to explore this idea by looking at paralogs and variants at identical sites across gene families, and we found evidence that was strong enough to be included in the official ACMG/AMP variant curation criteria.

An SCN1A mystery. A little more than a decade ago, I struggled to make sense of the disease history of a child with a de novo SCN1A variant. The clinical presentation was unusual and, while we had framed the clinical presentation as atypical Dravet Syndrome, there were many unusual features including episodes of dystonia and a much earlier seizure onset. This variant had never been seen before in any other individual and no other information was available.

Looking at paralogs. At some point during the treatment journey, we tried to gather more evidence by looking whether the same variant had ever been seen in other sodium channel genes. Sodium channel genes are very conserved, and we hoped that we could obtain more insight by looking at the larger gene family. This turned out to be the case – an identical amino acid exchange in SCN2A had been seen previously and was found to be a gain-of-function variant. We took the (at the time) risky step and treated our patient with sodium channel blockers and symptoms improved. What helped us make this decision? The fact that the variant was “paralogous” to a known gain-of-function variant in a closely related gene. But was this more than an educated guess, and is there a general principle behind this?

The paralog principle. In our study led by Tobias Brünger and Dennis Lal, we asked a deceptively simple question: if a variant is known to be pathogenic in one gene, what does it mean when we see a variant at the same aligned residue in its paralog: a member of a close related gene within a larger gene family? To answer this question, we mapped over 590,000 pathogenic and 1.9 million population variants across 9,928 genes grouped into 2,054 gene families. The result: incorporating information from paralogous variants increased the number of classifiable amino acid positions by more than five-fold. This provided strong statistical evidence that these echoes of pathogenicity matter.

A practical extension. But how does information about paralogous genes influence variant curation criteria? The current ACMG/AMP guidelines allow us to use prior evidence from the same gene if the same amino acid is altered (PS1) or a different substitution occurs at that site (PM5). But these criteria have traditionally ignored the rich information in paralogs. In our study by Brünger et al., we show that incorporating paralogous variants at conserved residues provides likelihood ratios for pathogenicity that meet or exceed existing thresholds. This information is already included in the updated ACMG/AMP guidelines that were released in January 2025 by our ClinGen group. In brief, paralogs are not just of interest academically; they are already included in the current variant curation criteria that every diagnostic laboratory uses.

What you need to know. Paralogous genes offer a powerful method for interpreting rare missense variants, especially in epilepsy-related ion channels. In our study, we show that pathogenic variants in one gene provide meaningful evidence for the pathogenicity of variants in a related gene at the same position, a principle that has since been included in the official ACMG/AMP variant curation criteria. This principle expands the number of classifiable residues five-fold and increases odds of pathogenicity thirteen-fold. Especially in highly conserved gene families like the sodium channels, these insights help us connect structure, function, and phenotype across genes, and bring us one step closer to closing the diagnostic gap.