ClawBio + Nebius · Impact Hub Berlin

Agentic
Genomics

Validated skills are necessary, and not sufficient.
ClawBio + Nebius Hackathon Berlin, Tuesday 18 August 2026, Impact Hub, Rollbergstrasse 28a
The shift

Two waves of LLMs in biology

First wave

Information retrieval, chat, summarisation.

Useful, but incremental.

Second wave

Autonomous execution, multi-step planning, tool use. Writes, debugs and runs code. Recovers from failure.

The bottleneck moves from producing code to validating results.

64% of this room uses coding agents daily. You already live in the second wave.

Definition

What makes a system agentic

01 Autonomy

Run-time decisions, not a static workflow.

02 Domain constraint

A library of validated skills, not ad hoc code.

03 Iterative refinement

Reads intermediate results, recovers from error.

04 NL mediation

You describe intent; the agent translates it to execution.

Jointly necessary, and falsifiable by the perturbation test: an agent that ignores a corrupted intermediate output was never agentic.

The new bottleneck

Silent failure, with a plausible face

Agents produce results faster than a human can check them.

AutoBA

Pipelines skipping critical steps; wrong tool for the data type.

Adv. Sci. 2024

Single-cell agents

Incomplete designs; inconsistent answers to identical queries.

Brief. Bioinform. 2025

Boiko et al.

Protocols syntactically correct, scientifically invalid, passing execution checks.

Nature 624, 570 (2023)

Our own audit

A ClawBio skill silently returned “all normal” for 51 drugs on an empty input file.

Independent audit, S. Kornilov

Framework

Validation proportional to consequence

Research grade

Hypothesis exploration. Unit tests, adversarial inputs, every output reviewed. False positives tolerable.

Benchmarked

Publishable analyses. Public references (GIAB, GeT-RM), independent benchmarking, published failure modes.

Clinical grade

Patient care. Multi-centre external validation, signed reproducibility packages, CLIA/CAP.

Today you build at tier one. Knowing which tier you are at is the skill.

The library under test

ClawBio

An agent-native skill library for bioinformatics. Open source, local-first, reproducible.

96
skills
1,105
GitHub stars
46
contributors
8,182
Galaxy tools bridged

github.com/ClawBio/ClawBio · MIT · and the artefact under test in everything that follows.

The empirical question

Can a plain-text skill specification reach clinical grade?

Pharmacogenomics. Genotype to phenotype to drug recommendation, with ground truth fixed by CPIC.

DPYD rs3918290 T/T plus standard fluorouracil can kill the patient.
13,200
evaluations
8
frontier models
110
CPIC Level A cases
5
configurations

Claude Opus 4.5 · Claude Sonnet 4.5 · GPT-5.2 · GPT-4.1 · o3 · o4-mini · Gemini 2.5 Flash · DeepSeek V3 · three replicates · 21 genes

Result 1

Where the correctness lives

62.4%
free generation
55.1%
RAG generation
97.0%
authored rules, generated
97.3%
authored rules, executed

The jump comes from authoring the decision logic, not from which model applies it.

Recommendation accuracy, 2,640 evaluations per configuration. Execution adds determinism, bounded failure and explicit abstention, not accuracy.

Result 2 · the uncomfortable one

Retrieval made it less safe

Recommendation accuracy and lethal-class errors across five configurations

Lethal-class errors, of 336 attempted per configuration. Free 54RAG 137 → authored rules 23–26. 74 of the 83 extra errors are on HLA loci; HLA-B*15:02 goes from 0 of 48 to 45 of 48.

Result 3 · the one that should worry you

Curated accuracy does not transfer to real genomes

Accuracy by cohort: curated benchmark versus four real cohorts

96.4% curated · 68.2% family WGS (n=4) · 60.6% Iberian (n=93) · 53.4% Peruvian · 37.2% East African (Uganda, 6,407 individuals). Same skill, same models. “Whose genome does this fail?” is not rhetorical.

Result 4

Two ways to be useless

The deterministic mapping

Never wrong. Often silent.

Answers only where its vocabulary holds the called diplotype. Correct on every answer it emitted, in all four cohorts.

The agent

Answers almost everything. Wrong on 21–41% of it.

Declines on 13.5% to 37.2% of states, and the rate of clinically wrong answers shows no cohort ordering at all.

A confident wrong answer is worse than an honest refusal.
Result 5 · practical

Give the model the definitions, and pick your model

Coverage and accuracy by model, with and without allele definitions

Supplying the caller's allele-definition table moved the best model from 39.6% to 96.7% against the caller. Eight models holding the same table span 46.4% to 81.4% against external consensus (GeT-RM), n=527 pairs per model. Structure helps every model; it does not make them equivalent.

The thesis

Validated skills are necessary and not sufficient

Authoring the decision logic takes recommendation accuracy from 62.4% to 97.0%. Executing it adds determinism, bounded failure and explicit abstention.

And it still reaches only 37.2% on East African genomes, still inherits every error in the input call, and still faithfully executes a tampered contract: 88 of 90 responses echoed deliberately corrupted rules, and none reverted to the correct answer.

Corpas, Iacoangeli, Bourdenx, Aldraimli, Jabalameli, Skene, Guio, Fatumo · Cell Genomics, submitted 16 August 2026 · Zenodo 10.5281/zenodo.21905387 · tag agentic-pgx-benchmark-v2.3

Today

The shape of what you build

Start deterministic

Run a ClawBio skill. It computes an answer you can check, and tells you what it could not call. That is your ground truth for the day.

Add the agent

Put a Token Factory model in charge of it, with Tavily for grounding. Now it chooses, runs, reads the output and reports.

Then go further

Call a hosted model as a tool, Boltz-2 or DiffDock. Chain skills. Or write a new skill and open a PR.

Any question you like: pick one of the three challenges, or bring your own and we scope it at 13:05. Public data, and being wrong has to matter.

A jury picks three winners on originality, impact, and how well you use Nebius and ClawBio. The room picks a fourth, the community prize. A confident answer with an invented source scores below an honest abstention.

Get building

The whole setup

1.  Link + token  →  #berlin-general on Slack
2.  Open the link, paste the token, Connect
3.  BioNeMo Research Agent, on Nemotron 3 Super

4.  Paste your challenge prompt:
    docs.clawbio.ai/hackathon/berlin/tracks

Nothing to install. You drive an agent, and it reads the skills, runs them and chains them. A grounded answer costs about 1,300 tokens and seven seconds, so your credits are not the constraint on what you attempt.

Ask it first what tools it actually has, then build to that. All the challenge data is ready at docs.clawbio.ai/hackathon/berlin/data and needs no account.

docs.clawbio.ai/hackathon/berlin · #berlin-help in Slack, all day · demos at 16:30

1 / 15
ClawBio + Nebius · Berlin 2026