Helena Bioinformatics — Undertone
Every association
has an undertone.
Undertone turns genetic associations into testable biological hypotheses. It combines GWAS, fine-mapping, regulatory, expression, and functional evidence to show which variant likely acts, in which cell type, on which gene, through which mechanism.
Drag to see what's underneath
A GWAS hit tells you a locus moved. It doesn't say which variant did it, in which cell, on which gene, or by what mechanism. Most tools stop at the association. Undertone keeps going, and shows its work at every step.The problem with a single confidence score
Real handoff · interface-shaped
Show the data we have. Show the gaps we do not.
The current TCF7L2 handoff is rendered without filling in missing science. Two source records become four allele-specific identities, while every absent artifact stays explicit.
Submission preview
TCF7L2 credible-set handoff
The email named only C>T and G>T. The supplied VCF contains ALT G,T and A,T. Undertone preserves the file and exposes all four identities rather than guessing which alleles were intended.
Credible-set membership is present, but the handoff has no fine-mapping method, posterior values, parameters, or source. Undertone must not manufacture a rank.
Submission package completeness
Present fields are not equivalent to a validated dossier.
Illustrative enriched dossier
Enter the relationships, not just the result.
Select a candidate variant. Follow its regulatory path. Open any node to see what supports it, where it came from, and what remains uncertain.
Mechanism map
rs4506565
chr10:112,996,282 C>TThe risk allele may reduce islet enhancer activity and alter TCF7L2 regulation in pancreatic beta cells.
Uncertainty stays visible
The proposed mechanism lacks direct perturbation evidence.
Undertone carries the gap forward and identifies the experiment that could resolve it instead of hiding it behind a score.
The evidence model
From locus to hypothesis, one layer at a time.
Each layer narrows the field. What's known stays separate from what's inferred, so the hypothesis at the bottom is only as strong as the evidence stacked above it.
A locus crosses genome-wide significance in one or more cohorts.
Statistical fine-mapping narrows the signal to a credible set of candidate variants.
Chromatin state and transcription factor footprints flag which candidate sits in active regulatory sequence.
eQTL and allele-specific expression data link the variant to a candidate target gene.
Reporter assays, MPRA, or CRISPR perturbation would confirm the variant changes activity. Most loci don't have this yet, and Undertone says so plainly.
One stated hypothesis: this variant, in this cell type, acts on this gene, through this mechanism, traceable back to every layer above.
Example dossier
What a variant's file actually looks like.
A working example. Each row keeps its evidence type visible, so nothing observed gets mistaken for something predicted.
Type 2 diabetes, pancreatic islet
Method
How a locus becomes a hypothesis.
Ingest
Pull GWAS, fine-mapping, regulatory, and expression evidence for a locus from public and licensed sources.
Separate
Split observed data from computational prediction, and flag where sources disagree.
Map
Trace each candidate variant to its likely target gene and cell context.
Hypothesize
Assemble one testable mechanism statement, with every claim traceable to its evidence.
Flag gaps
Mark exactly what still needs experimental validation, instead of hiding it behind a score.
Stop trusting the surface color.
Undertone is in early access for research groups working on GWAS follow-up and variant-to-function studies.
Request access