The only tool that writes your multi-ancestry Results section
Upload GWAS summary statistics. CohortLens generates per-ancestry Manhattan plots, admixture-aware LD panels, and Results prose that correctly describes EUR, AFR, and admixed findings — language that PIs at diverse-cohort institutions currently write by hand.
Also works for single-ancestry cohorts. No individual-level data, no IRB exposure.
No credit card required · Free tier includes 5 cohorts · ORCID or Google sign-in
Built for the way genomics research actually works
Every design decision was made against one constraint: no individual-level data, ever. That means it works at the summary-stats stage — the data you share with collaborators, post as a preprint, or send to reviewers.
IRB-safe by design
Summary statistics only — no VCFs, no individual-level genotypes, no PHI. Every output is citable in a Methods section. Deploy on-prem for complete institutional control.
Writes your multi-ancestry Results
No other tool generates Results prose for admixed cohorts. Per-ancestry thresholds, RFMix tract interpretation, and cross-stratum concordance language — all grounded in your actual summary statistics.
AI narrative grounded in data
Claude generates Results prose and figure legends from your actual significant hits. Every statistic cited must be present in the input — no hallucinated p-values, ever.
Shareable like Figma
Every cohort gets a collaborative URL. Annotations, per-figure comments, side-by-side cohort comparison, and public demo links — no email attachments, no static PDFs.
Results prose grounded in your actual hits
Claude reads your significant associations, fine-mapping PIPs, and colocalization posteriors — then writes a Results paragraph that names the real SNPs, real effect sizes, and real genes. Every number cited must exist in your input file.
- AJHG · Nature Genetics · PLOS Genetics · HMG journal styles
- Figure legends generated alongside Results prose
- Cross-domain narrative linking GWAS → eQTL → RNA-seq
- No hallucinated statistics — [BRACKETS] for anything unknown
We identified a genome-wide significant association at chromosome 11 (rs123456; β = 0.18, SE = 0.03, p = 2.1 × 10⁻⁹, N = 15,000) in our EUR–AFR admixed gestational-age cohort. Fine-mapping identified one credible set of three variants (95% CS), with rs123456 achieving the highest posterior inclusion probability (PIP = 0.94). Colocalization analysis with GTEx v10 uterine eQTLs revealed strong evidence of a shared causal variant with PGR expression (PP.H4 = 0.903), consistent with a progesterone-receptor-mediated effect on gestational duration.
From upload to submission ZIP in minutes
Upload summary statistics
GWAS, RNA-seq DE results, COLOC output, GCTA-COJO, RFMix — any format. Files are relayed through your own account's secure storage. No individual-level data is ever accepted.
Generate interactive figures
D3-powered Manhattan plots, volcano plots, ancestry-stratified views, colocalization bars — all rendered in your browser, shareable with a link.
Export publication-ready output
Results prose, figure legends, and a ZIP with Word manuscript + figure SVGs pre-formatted for AJHG, Nature Genetics, PLOS Genetics, or HMG journal styles.
13 omics domains, one platform
Each domain has its own parser, figure handler, and narrative context module. Upload the output of any standard tool — no reformatting required.
Results prose for admixed and multi-ancestry cohorts
PIs at UCSF, UPenn, Vanderbilt BioVU, and similar institutions spend real time manually crafting Results language for multi-ancestry GWAS. CohortLens generates it — per-stratum effect sizes, correct significance thresholds per ancestry, RFMix tract interpretation, and cross-ancestry concordance prose — all grounded in your actual summary statistics.
EUR-ancestry stratum (N = 8,412): rs123456 reached genome-wide significance (β = 0.21, SE = 0.04, p = 1.8 × 10⁻⁸). LD r² coloring against the CEU 1KG reference panel reveals a tight credible set spanning 42 kb at 11q22.1.
AFR-ancestry stratum (N = 6,231): directionally consistent but attenuated (β = 0.09, SE = 0.05, p = 0.04). Local ancestry tract analysis (RFMix) confirms EUR haplotype background at this locus in AFR-ancestry carriers.
EUR–AFR admixed (N = 357): per-ancestry thresholds applied (p < 3.8 × 10⁻⁸ EUR; p < 5.0 × 10⁻⁸ AFR). Effect sizes intermediate and directionally concordant across all three strata — supporting a shared causal locus.
Per-ancestry thresholds
Genome-wide significance thresholds adjusted per ancestry group — not one size fits all.
RFMix tract interpretation
Uploads .msp.tsv output, aggregates to IRB-safe window proportions, interprets tracts in prose.
Admixture-aware LD panels
EUR-AFR → CEU + YRI side-by-side. No blended r² — every value is a real single-population number.
Cross-stratum concordance
Narrative explicitly addresses directional concordance and effect size differences across ancestries.
Works for single-ancestry cohorts too — ancestry stratification is opt-in, not required.
Start free, upgrade when you publish
- 5 cohorts
- All 13 omics domains
- Interactive D3 figures
- Public share links
- Collaborative annotations
- Unlimited cohorts
- AI narrative generation (Claude)
- Preprint-ready ZIP export
- 300 DPI TIFF + journal presets
- Journal compliance checking
- Reviewer response generator
- PowerPoint export
- Priority support
from $599/mo · on-prem available
- Everything in Pro
- Multi-seat org billing
- On-prem Docker deployment
- Summary stats never leave your network
- One-command setup script
- Dedicated support channel
hello@stellagenomics.com
Need Stella grant intelligence too? StellaGenomics Suite — $119/mo · On-prem: one-command setup via scripts/on-prem-setup.sh
Stop writing your multi-ancestry Results by hand
Upload GWAS summary statistics and generate ancestry-stratified figures plus a Claude-drafted Results section in under five minutes. No individual data required.