Disease-Associated Non-Coding Variants: From Risk Locus to Cellular Phenotype

Fine-mapping a locus produces a credible set. Editing a variant produces an isogenic pair. Both are technically tractable. The step in between — deciding what to measure — is where most projects stall, because the phenotype is not given by the genetics.

A risk variant in a microglial enhancer does not come with a readout. The investigator must choose one: baseline expression, stimulated cytokine output, phagocytic capacity, or a transcriptional programme. Each is a different claim, with a different sensitivity, and only some will show the effect.

This article is about choosing that endpoint deliberately. The statistical path from GWAS signal to candidate variant is covered in the cis-regulatory cluster; here the subject is the cellular model and the phenotype.

The first two rows usually suffice to pick a lineage; none of them establishes the phenotype. State the basis for the cell-type choice — reviewers in this field ask for it.

Five phenotype classes, in increasing specificity:

The second row is the most underrated: in a heterozygous background, allele-specific expression compares the two alleles within the same cell, removing most confounders. Where the cell type permits it, design for heterozygosity rather than editing both alleles.

• • Both edit directions where the cell system allows: risk allele introduced into the protective background, and the reverse. Concordance across directions is markedly stronger than one direction alone.
• • Multiple clones per genotype. Clonal variation in iPSC-derived cells is large; a single pair may not resolve a modest effect.
• • Matched differentiation batches. Edit, then differentiate clones in parallel — differentiation batch effects can exceed the variant effect.
• • Stimulation as part of the design. Many non-coding variants act on the response to a stimulus rather than the baseline; pre-specify the stimulus.

Non-coding variant effects are small. A 10–30% change in expression is a typical finding and a meaningful one, provided it is reproducible and directionally consistent with the genetic association.

Practical implications:

• • Power the experiment for a small effect, not for the effect you hope to see.
• • Report confidence intervals, not only fold changes.
• • Compare to the population effect size where an eQTL exists; a 25% cellular change is plausible, a fivefold change usually indicates the model is measuring something else.

EDITGENE builds isogenic variant panels in iPSC-derived and primary-like cells, with heterozygous backgrounds for allele-specific readouts, validated differentiation, stimulated functional assays, and matched batch replication.

Tell us the locus, the credible set, and the cell type you suspect. Contact EDITGENE to request a project consultation or quote.

Cell-type assignment and phenotype choice are inferences from available data; both should be stated explicitly, along with the evidence supporting them.

Cellular effect sizes need not match population effect sizes, and findings in a model system require corroboration before a disease mechanism is claimed.

• • Nasser J, et al. Genome-wide enhancer maps link risk variants to disease genes. Nature. 2021;593:238–243.
• • Ulirsch JC, et al. Interrogation of human hematopoiesis at single-cell and single-variant resolution. Nature Genetics. 2019;51:683–693.
• • Corces MR, et al. Single-cell epigenomic analyses implicate candidate causal variants at inherited risk loci for Alzheimer's and Parkinson's diseases. Nature Genetics. 2020;52:1158–1168.
• • Wang QS, et al. Genetic and epigenetic effects of non-coding variants. Nature Reviews Genetics. 2023;24:1–18.