SCORIAL
Autonomous research engineJoin waitlist
Computational biology · research engine

Ask a research question. Get a submission-ready paper.

SCORIAL is an autonomous research engine for computational biology. Describe what you want to know in plain English — it finds real public data, runs the actual analysis, generates the figures, and writes up what it found as a submission-ready report, with a reproducible manifest for every run.

Private beta for computational biologists · no spam

How a run works

One question in. A complete, reproducible analysis out.

Each stage runs autonomously and hands off to the next. The order is the pipeline — and every stage can stop the run cleanly if it can’t proceed honestly.

01Ask

Pose the question in plain English. No pipeline to configure, no dataset to pick by hand.

02Find data

Resolves a real public dataset from GEO that actually fits the question — or halts if none does.

03Run

Executes the actual analysis — differential expression, enrichment, clustering — not a description of one.

04Figures

Generates the plots directly from the computed results: volcano, PCA, dotplots, UMAP.

05Write up

Drafts the findings grounded in what it actually computed — the numbers come from the run.

06Manifest

Emits a reproducible record: dataset, design, method, versions, and checksums for the whole run.

Real data · reproducible

It runs on real public datasets — and the same question gives the same answer.

Real data, not invented IDs

SCORIAL resolves and analyzes actual GEO datasets. It never fabricates an accession or fills a gap with a plausible-looking dataset that doesn’t exist.

Deterministic, bit-for-bit

Given the same question and data, a run reproduces exactly — down to the significant-gene counts — because the analysis is pinned, not improvised.

A receipt for every run

The manifest records the dataset, the design it derived, the method it ran, and the checksums — so anyone can trace or re-run the result.

Honest-halt

When it can’t do something reliably, it stops and says so.

The honest-halt is SCORIAL’s defining behavior. If a step can’t be completed reliably — the data isn’t raw counts, the design is ambiguous, no dataset fits the question — it stops cleanly and tells you exactly why. It stops the moment it can’t verify a step — it never guesses forward.

A wrong answer, delivered confidently, is worse than no answer. SCORIAL is built to prefer the honest stop.

Proof

A general chatbot describes the analysis. SCORIAL runs it.

Put the same question to each. One hands back text you still have to run and trust yourself. The other executes it end to end — and leaves a reproducible receipt.

Capability
General chatbot
SCORIAL
Runs the analysis on real data
General chatbotDescribes or guesses at it
SCORIALActually executes it
Uses verified real datasets
General chatbotMay invent dataset IDs
SCORIALResolves real GEO data
Same question → same answer
General chatbotVaries each time
SCORIALDeterministic, bit-for-bit
When it can’t do something
General chatbotGuesses confidently
SCORIALHonest-halts and says why
Reproducible receipt
General chatbotNone
SCORIALFull manifest
End to end
General chatbotGives you code to run yourself
SCORIALDoes the whole thing
Early access

Put a question to the engine.

SCORIAL is in private beta for computational biologists. Join the waitlist and we’ll reach out as seats open.