Ready to get started?

Cohort analysis, in a conversation.

Run GWAS, PRS, fine-mapping, and ancestry-stratified analyses across thousands of samples. Get publication-ready figures and code that's reproducible by your reviewers.

Discovery · Simple, Fast, Accurate

See our publications

bystro.io/professional

Bioinformatics

“To have it actually generate, on the fly, pages of fully valid statistical analysis code, execute that code locally for security, generate volcano plots, and interpret the results was astonishing. This will be transformative for the field.”

Charles DeCarli, MD

Distinguished Professor of Neurology, UC Davis · Co-Director, Alzheimer's Disease Research Center

200+

annotation fields

23+

population frequencies

1000s

samples per job

100%

cited & reproducible

Move beyond the pipeline into the answer.

One platform. Configured for how your team works.

Research institutions

Publication-grade analysis, without the bioinformatics overhead.

THiNK reasons directly from your annotated variants across 200+ fields from ClinVar, gnomAD, CADD, SpliceAI, and more, then pulls live literature to support every finding. The kind of depth that used to take a dedicated analyst now lives in a conversation.

Run GWAS, PCA, survival analysis, and cohort comparisons in plain language
Methods text generated alongside results, written to survive peer review
Every finding cited to peer-reviewed sources and specific database versions
In reproducing published studies, Bystro has surfaced errors that passed original peer review

Example queries

THiNK · agent · running

"Run a GWAS on our 2,000-sample IBD cohort, correct for population stratification, and generate a Manhattan plot ready for submission."

THiNK · agent · running

"Compare variant burden across three disease subtypes and write the methods section describing the statistical approach."

Every answer is grounded in your variants and cited to peer-reviewed sources.

Research Use Only. This platform is for research, educational, and informational purposes. Not intended for medical diagnosis, treatment, or clinical use. Always consult qualified healthcare professionals.

Drug development

Biomarker discovery that moves at the speed of the science.

From VCF upload to annotated, stratified cohort, DASH processes petabyte-scale datasets with 200+ annotation fields, ancestry-corrected polygenic risk scores, and automatic QC. THiNK reasons on top of it to surface what matters for your program.

Stratify patient populations by variant burden, ancestry, or polygenic risk before trial enrollment
Identify rare loss-of-function variants across biobanks and rank by predicted effect size
Version-tracked annotation surfaces data and statistical inconsistencies early, before they reach regulatory review
Auditable, versioned outputs ready for regulatory and BD review

Example queries

THiNK · agent · running

"Identify genetic subgroups in our Phase II trial cohort that correlate with treatment response and visualize the variant landscape."

THiNK · agent · running

"Flag rare loss-of-function variants in the PCSK9 pathway across our 50,000-sample biobank and rank by predicted effect size."

Every answer is grounded in your variants and cited to peer-reviewed sources.

Clinical research

Every variant interpretation, fully traceable.

THiNK gives clinical research teams the genomic depth of a full bioinformatics group, with plain-language summaries, cited evidence for every finding, and outputs that hold up under scrutiny.

Variant interpretations grounded in your annotated data, not general literature synthesis
Cohort-level reporting across patient populations with statistical context
Plain-language summaries alongside full technical outputs
De-identified data workflows with enterprise data agreements

Example queries

THiNK · agent · running

"Summarize the variant profiles of our cardiology cohort and flag patients with elevated polygenic risk scores for follow-up."

THiNK · agent · running

"Generate a de-identified report comparing variant burden between our responder and non-responder groups with statistical significance."

Every answer is grounded in your variants and cited to peer-reviewed sources.

Biobanks & population cohorts

Annotation at biobank scale, without the weeks-long queue.

Multi-hundred-thousand-sample datasets and petabytes of sequencing files strain even well-resourced operations, and bioinformatics talent remains scarce. In a peer-reviewed Genome Biology benchmark, Bystro annotated 84.9M variants across 2,504 genomes in under 3 hours, with runtime that scales linearly with cohort size — running comfortably on premises, so sensitive genomic data never has to leave your environment.

Runs on premises, not a cluster of parallel instances — your genomic data stays inside your own infrastructure
Annotation, QC, natural-language search, ancestry, and PRS in one workflow, no pipeline to assemble
Ancestry inferred automatically and matched to PRS calculations, closing the portability gap in European-derived scores
Built-in PRS for common traits, plus custom GWAS Catalog scores, ancestry-adjusted by default

Example queries

THiNK · agent · running

"Harmonize our genomic, transcriptomic, and proteomic datasets and flag which multiomic signatures associate most strongly with cognitive decline."

THiNK · agent · running

"Please create a cohort of 800 samples with ancestry-corrected polygenic risk scores for type 2 diabetes across our full biobank and flag participants above the 90th percentile within their ancestry group."

Every answer is grounded in your variants and cited to peer-reviewed sources.

The annotation engine and the agentic AI that reasons on top of it.

DASH produces the grounded data layer: annotated, versioned, and reproducible. THiNK reasons on top of it. Together they give your team the complete stack from variant to answer.

The first end-to-end genomic agentic AI designed from the ground up by geneticists. Every genetic finding traces directly to your variants. No black boxes. No guesses.

Deterministic

DASH

The grounded data layer: annotated, versioned, and reproducible.

Every one of 200+ annotation fields indexed and filterable

Automatic QC every job: Ti/Tv, het/hom, missingness, Watterson's theta

Every annotation traceable to a specific database version

Arrow Feather v2 dosage matrices, ML-ready the moment annotation completes

CADD PHRED scores, SpliceAI and Pangolin splice predictions, PolyPhen-2 and SIFT missense impact scores, and PhyloP evolutionary conservation.

Reasoning

THiNK

Runs real analysis on that layer, not just reasoning about it.

Natural-language variant search — no query language, no SQL

Cross-ancestry polygenic risk scores from any GWAS study

Executes Python in a persistent sandbox: GWAS, PCA, regression, clustering

Publication-quality figures embedded in the conversation

Runs auditable analyses, and verifies every claim. Results and publication-quality figures embedded in the conversation.

Annotation databases

ClinVargnomAD v4dbSNPCADDSpliceAIPhyloPPharmGKBOMIMRefSeqUniProtCOSMICCustom annotations

Deployment

Private servers, on-premises available.

On-premises

Your hardware, your control

Cloud

Secure private cloud, fastest start

Used by

Emory School of Medicine

Department of Neurology

Talk to our team.

Enterprise deployments, on-prem installations, custom data agreements, and academic licensing. Tell us about your use case and we'll figure out the right fit.

Custom data agreements & BAAs

On-prem & private VPC deployment

Academic & non-profit licensing

White-label options for CROs

team@bystro.io