Closing the Interpretation Gap for Variants of Uncertain Significance

You are reading version 1 of this direction, frozen 2026-10-04. The latest version is version 2.

Version 1 of 2145 papers177 conceptsfrozen 2026-10-04

The cost of reading a human genome has plummeted, but our ability to understand what we are reading has not kept pace. When a clinician sequences a patient’s DNA, they often find a "Variant of Uncertain Significance" (VUS)—a genetic typo that might cause a devastating disease or might be completely harmless. This "interpretation gap" is the primary bottleneck in precision medicine.

Recent research has shifted from simply cataloging these variants to building functional and computational frameworks that can predict their impact before a patient ever walks into a clinic.

The Functional Stress Test

The most direct way to understand a mutation is to watch it work. Traditional genetics waited for a patient to appear with a symptom, but researchers like Zsofia Kote‐jarai and Lucy Matthews pioneered a "stress test" approach. In Accurate Prediction of BRCA1 and BRCA2 Heterozygous Genotype Using Expression Profiling after Induced DNA Damage, they demonstrated that by exposing cells to ionizing radiation, they could force a hidden genetic weakness to reveal itself. The cells' transcriptional response to the damage acted as a functional readout, distinguishing pathogenic mutations from benign ones with near-perfect accuracy.

This concept has evolved into high-throughput "Cell Painting" assays. John Arevalo and Juan C. Caicedo developed cmVIP (cell morphology-based Variant Impact Phenotyping), which uses deep learning to analyze the physical structure of a cell. As described in Cell Painting predicts impact of lung cancer variants, a mutation that disrupts a protein's function often leaves a "morphological signature"—a subtle change in the shape of the nucleus or the arrangement of organelles—that AI can detect more reliably than a human observer.

Pathogenicity as a Misfolding Problem

Why do these mutations cause trouble? Often, it isn't because the mutation "breaks" the protein's active site. Instead, the mutation makes the protein slightly unstable or "floppy."

Sofie V. Nielsen and Amelie Stein have shown that for proteins like MSH2 (linked to Lynch syndrome), the cell’s own quality control machinery is the culprit. In Predicting the impact of Lynch syndrome-causing missense mutations from structural calculations, they argue that most disease-causing variants trigger the proteasome to degrade the protein prematurely. Similarly, Lene Clausen found that Birt-Hogg-Dubé syndrome is essentially a protein misfolding disease, where the cell treats functional but slightly misshapen proteins as "trash."

Learning the Grammar of Evolution

If we cannot test every mutation in a lab, we can look at the "experiments" evolution has already performed over millions of years. If a specific amino acid has remained unchanged from fish to humans, it is likely critical.

Jonathan Frazer and Yarin Gal developed EVE (Evolutionary model of Variant Effect), a deep generative model that learns the "grammar" of protein sequences. By training on massive cross-species datasets, EVE can predict the clinical significance of a human variant without ever seeing a single piece of human clinical data. As detailed in Large-scale clinical interpretation of genetic variants using evolutionary data and deep learning, this unsupervised approach avoids the biases of clinical labels and matches the accuracy of expensive experimental assays.

Other models like MVP, developed by Hongjian Qi, use residual networks to distinguish rare pathogenic mutations from neutral background noise, providing a computational shortcut to predicting missense variant pathogenicity.

Working Backward from the Phenotype

Sometimes the answer isn't in the cell or the sequence, but in the patient's face. Alexander J.M. Dingemans and colleagues introduced PhenoScore, a framework that combines facial recognition AI with clinical symptom analysis. In PhenoScore: AI-based phenomics to quantify rare disease and genetic variation, they demonstrate that the "visual gestalt" of a patient can provide the missing evidence needed to reclassify a VUS. If a patient’s physical features match the known signature of a syndrome, the uncertain variant in that gene becomes the likely smoking gun.

The Path Forward: Multiplexed Maps

The ultimate goal is to move from reactive testing to proactive mapping. Projects like MaveDB v2, led by Alan F. Rubin, are aggregating "Multiplexed Assays of Variant Effect" (MAVEs). These assays test thousands of mutations simultaneously, creating a "lookup table" for clinicians.

By combining these comprehensive maps (like the one created for PSAT1 by Michael J. Xie) with evolutionary models and deep phenotyping, the research community is systematically turning the "uncertain" into the "actionable."

Papers in this version

Showing 60 of 145 papers.

History & citationVersion 2 · 2026-10-04

Current · Permanent link

  1. Version 2 (latest) · 2026-10-04 · 136 papers
  2. Version 1 · 2026-10-04 · 145 papers
Cite this direction

Latest version (a living page):

@misc{lacuna_variants_of_uncertain_significance_d18c4126,
  title = {variants of uncertain significance},
  author = {{Lacuna}},
  howpublished = {Lacuna research direction},
  year = {2026},
  url = {https://lacuna.tiptreesystems.com/direction/variants-of-uncertain-significance/rdr_d18c412600a84295b18f500db7df208a},
  note = {Latest version of a living research direction; cite a version for a fixed text. Accessed 2026-11-21},
}

This exact version (version 1, frozen text):

@misc{lacuna_closing_the_interpretation_gap_for_varia_d18c4126_v1,
  title = {Closing the Interpretation Gap for Variants of Uncertain Significance},
  author = {{Lacuna}},
  howpublished = {Lacuna research direction},
  year = {2026},
  url = {https://lacuna.tiptreesystems.com/direction/closing-the-interpretation-gap-for-variants-of-uncertain-significance/rdr_d18c412600a84295b18f500db7df208a/v/1},
  note = {Version 1, frozen 2026-10-04. Accessed 2026-11-21},
}