evidnaEARLY RESEARCH

AN OPEN LAB FOR CURIOUS MINDS

Life is complex.
Start with
one question.

Meet Evidna, an open-source biology lab that pairs real DNA with AI reasoning. Read a sequence. Follow a mutation. Let the evidence lead.

PYTHON + BIOPYTHON / OPEN SOURCE / EVIDENCE FIRST

SPECIMEN 001HOMO SAPIENS / HBB
One letter. A different protein.
ATG GTG CAC CTG ACT CCT GAG5′ → 3′
LESS GUESSWORK. MORE EVIDENCE.

From biological data to questions you can actually test.

01 / THE FIRST STUDY

A SMALL CHANGE, A BIG QUESTION

Meet your first gene.

HBB encodes beta-globin, part of the protein that carries oxygen in blood. Explore how a single DNA substitution changes its encoded amino acid.

EVIDNA / SEQUENCE WORKBENCHPYTHON-COMPUTED DATA
DNA → CODON → PROTEINLoading

REFERENCE

GAG

Glutamate E

SUBSTITUTION

GTG

Valine V

01

A change in the recipe.

Loading results from the Biopython analysis…

Calculated mass change
—
Estimated charge change · pH 7
—

Unprocessed chain. Average mass. No heme or physical measurement.

A sequence change is an observation. Its effect on a living system needs more evidence.

Interactive sequence workbench. Results generated by the same Python engine you run locally.

RESULTS, WITH THEIR LIMITS

What the lab found.

Repeat runs. The same public inputs. The same results. Reproducibility is a first check, not a new discovery.

1,332single-base substitutions scanned
256 / 261UniProt protein annotations reproduced
9 / 9known sequence cases pass
3 / 3repeat runs agree

One gene. Six outcomes.

Predicted coding consequences. They do not establish severity, protein function, or clinical risk.

THE INTERESTING EXCEPTION

DNA is only part
of the story.

Two annotations sit outside a single-base explanation. One is Hb Bristol: published research reports a DNA change encoding methionine, followed by protein processing that yields aspartate.

A sequence-only model cannot explain that final protein by translation alone. The mismatch points to the model's boundary.

Read the original study ↗

Established biology from 1996. Evidna reproduced a known limitation; it did not discover this mechanism.

THE SCIENTIFIC METHOD, IN THE LOOP

A hypothesis is
a beginning.
Evidence is the work.

AI can help ask a sharper question. It cannot turn a plausible answer into a fact. Evidna keeps the distinction visible.

Read the technical whitepaper →
  1. 01

    Observe something real.

    Start with a public reference sequence. Keep its accession, source, and file fingerprint alongside every result.

  2. 02

    Ask. Then challenge.

    A researcher proposes a falsifiable hypothesis. A second AI pass challenges its assumptions and missing evidence.

  3. 03

    Keep an honest notebook.

    Attach an observation and its source. Revisit the hypothesis with that evidence. Keep the old version, too.

YOUR NEXT QUESTION STARTS HERE

A biology lab.
On your Mac.

Inspect the code. Reproduce the result. Learn what each step means.

Get the source ↗

FIRST EXPERIMENT / IN SILICO

# Clone the lab
git clone https://github.com/nulljosh/evidna
cd evidna

# Read the DNA. Trace the change.
uv run evidna.py analyze

# Ask the researcher and skeptic*
uv run evidna.py hypothesize

* Requires a signed-in Claude CLI. Two sequential Haiku calls. Public analysis is sent to the model.

THE HORIZON

Computational biology today. Reviewed experiments and instrument integration in the future. No physical experiments or autonomous hardware control in this release.