AI for Science

From biological questions to computational workflows.

We’re building a shared platform where scientists and engineers define biological problems and develop reproducible analyses.

Talk to us What we do

A model ranks a hundred variants.

A142V scores 0.91. The next one, 0.90. Why?

We keep the reasons with the result.

So your team knows why, before the bench.

  • Trainingwhat 0.91 was optimised for
  • Assumptionswhat the score takes for granted
  • Your assayevidence the model never saw

Predictions are cheap. Decisions are not.

One example: someone still has to work out why one candidate ranks above another.

A ranked list, as it arrives at the lab

RankVariantModel score
15L72F/A142V0.94
16D88N0.93
17A142V0.91
18A142V/T201S0.90
19G57S0.90

81 more below

Before testing number 17Where the answer is
  1. What does 0.91 measure?The training code
  2. What did the model see?The preprocessing
  3. What does it assume?Someone’s memory
  4. Does it fit our assay?Your protocol
  5. Why 17 and not 18?Nowhere. 0.01 apart.

The answers exist. They don’t travel with the score.

A prediction is one step in a loop.

Before it: a question, data and assumptions. After it: an experiment and new evidence.

We work from the question to the analysis.

A cycle: question, data, analysis, interpretation, experiment, evidence, and back to the question. TenesimBio works from the question through the interpretation. Evidence feeds the next question

What we do

From the question to software your team owns.

  1. Understand the science

    Agree on the question, what the data mean and which assumptions matter.

  2. Build the analysis

    Turn that shared understanding into a tested workflow, with appropriate explainability methods and clear limitations.

Our research combines specialised LLMs and explainable AI for biomolecular science.

How a project runs

No plan needed.

Bring what you have. We start from there.

  • A biological question

    Which candidates should we test next?

  • A dataset

    Is this enough data for a model?

  • An existing model

    Can we trust it on our targets?

  • A stuck workflow

    Can this notebook become a pipeline?

Articles

Notes on biomolecular AI and interpretability.

All articles

What are you working on?

A few lines is enough.

Email info@tenesimbio.com Meet the team