We’ve been building a new tool for the annoying stage of variant analysis where you’ve already run the usual annotations, filtered on frequency, checked ClinVar, looked at AlphaMissense/CADD/REVEL/phyloP, and you still have a handful of missense variants that all look vaguely plausible.

The tool lets you ask a more literal question:

Has this exact amino-acid substitution already occurred somewhere in primate evolution?

For a human missense variant, you can see:

whether the exact alternate amino acid is present in another primate which genera carry it whether it appears in one lineage or multiple separated lineages whether the residue is otherwise invariant across the primate tree whether nearby residues are variable or unusually constrained what is happening at the codon level underneath the protein alignment

That last part has been especially useful. A residue can look “conserved” in a standard track, but the surrounding region may be highly permissive. Or the amino acid may be fixed while the codon itself has accumulated synonymous change. Those are different evolutionary histories, even if they collapse to the same conservation score.

The use case is pretty simple:

deprioritize variants where the exact alternate has recurred across primates flag variants at residues that stayed fixed while neighboring positions changed sort out disagreements between predictors decide which variants are actually worth a functional experiment inspect the underlying biology instead of accepting one more opaque score

The resource covers 19,244 human genes and uses exome observations from 55 non-human primate genera mapped to GRCh38.

We’ve also run it against curated ClinGen and BRCA variant sets. The broad pattern is that expert-classified pathogenic variants are overwhelmingly absent from the primate observations, while subsets of benign, likely benign, and uncertain variants contain substitutions already observed in primates. We are not treating that as automatic classification, just as an extra piece of empirical evidence for ranking and review.

We’re looking for a few people who have a real gene, small panel, or missense shortlist they’d be willing to run through it.

Good test cases would be:

variants where predictors disagree rare disease candidates that survive standard filtering variants queued for MAVE, CRISPR, or other functional work genes with poorly characterized regions cases where a chimp-only comparison is not giving enough context

The interesting outcome is not “did the tool agree with us?” It’s whether the primate pattern actually changes what you would review or test next.

Disclosure: I’m affiliated with the team developing CodeXome. This is a research-use tool, and we’re specifically looking for honest feedback on where it adds signal and where it does not.

Reply here or DM me if you have a gene or variant set you’d like to try.