Introduction
What Geneticists call an allelic series is when one gene shows up in people at several different strengths, from switched off to overactive, with the health outcome tracking the strength in a consistent direction. It is the most convincing evidence human genetics can offer that a drug target is real.
Very few allelic series are supposed to exist, and the same handful of genes gets cited every time the subject comes up. Our new white paper argues that the examples were there all along, waiting on data that has only recently arrived.
A drug trial without a drug
Nobody had to run a dose-ranging study to know which way to push PCSK9, or how hard, or whether it was safe. Nature had already run it, in people, and the answer was sitting in the data.
The thing that predicts approval
About one in ten drugs that reach clinical testing is eventually approved. Most of those failures are good molecules aimed at the wrong gene.
The obvious assumption is that a bigger genetic signal buys more of that advantage. Minikel and colleagues checked, and it does not. Neither the size of the genetic effect nor how common the variant is makes much difference to the odds.
What does make a difference is being sure you have the right gene. A typical genetic signal falls in a stretch of DNA holding several genes, any of which could be the real culprit, and backing the wrong neighbor is the most expensive mistake in the field. An allelic series settles it: the same gene, hit independently several times, moving the same outcome every time. That certainty is what you are buying.
Interactive feature: trials nature already ran
Pick a condition in the interactive chart below. In each one, a single gene has been unintentionally "dosed" across a range of strengths in the human population, from switched off to overactive. Step through the groups and read what a drug developer learns from each.
How many more of these genes are already in the data?
The conditions above were found one at a time, over years. Once you can search for the pattern instead of stumbling onto it, the question stops being philosophical and turns into a ranked list.
Five things that change the math
If this evidence is so valuable, the fair question is why the field only has a handful of examples? The answer is that until very recently, the data needed to find them did not exist.
1. Enough people, sequenced widely enough
UK Biobank has now sequenced all 500,000 of its participants, and All of Us is heading for a million with far broader ancestry than earlier studies had.
Why this is meaningful: Many of the variants that complete a range exist in only one population. Breadth decides whether you find a group of people at all, or never learn they were there.
2. Measurements fine enough to see a gradient
Thousands of proteins and metabolites are now measured on the same participants, alongside their medical records.
Why this is meaningful: A graded effect needs a graded readout. A diagnosis code in a chart tells you yes or no. It cannot tell you how much, which is the entire point.
3. The right resolution
A variant that changes how much of a gene gets made can now be traced to a single gene in a single cell type, instead of averaged across a whole tissue.
Why this is meaningful: This is how a signal that sat misfiled under a neighboring gene for two decades finally landed on the right one, as it did for ulcerative colitis.
4. Variants tested by the thousand
A single experiment can now measure the effect of every possible variant in a gene, including severe ones no living person carries.
Why this is meaningful: The blank ends of a range get filled in deliberately rather than waited for. Nature no longer has to have run the experiment first.
5. Methods that hunt for the pattern
New statistical frameworks combine rare and common variants in the same analysis and search the whole genome for dose-response at once.
Why this is meaningful: Every example above was found by people who went looking at one gene. A pipeline that scans thousands at a time turns a run of lucky discoveries into a list you can work through.
PCSK9 took a decade and a small number of labs. The next several hundred genes with an equally clean story are already in view.
Download the white paper.
MAT-US-BBIO-0312 V1
Companion to The Coming Wave of Allelic Series: Why Genetics' Rarest Form of Evidence Is About to Become Common , BridgeBio Pharma and GondolaBio, August 2026. Findings drawn from Abifadel et al. (2003), Cohen et al. (2006), Diogo et al. (2015), Deaton et al. (2022), Akbari et al. (2022), Chang et al. (2025), Howard et al. (2026), Alegbe et al. (2026) and Stefanucci et al. (2026), all cited in full in the paper.
Approval dates are not from the paper and are taken from the U.S. Food and Drug Administration: alirocumab (Praluent) and evolocumab (Repatha), the first PCSK9 inhibitors, both approved in 2015 (Drugs@FDA2; FDA3); deucravacitinib (Sotyktu) approved 9 September 2022 for moderate-to-severe plaque psoriasis (Drugs@FDA, NDA 2149584). Laboratory locations for the 2003 and 2006 studies are the author affiliations given in those papers.