Genolator Lets Scientists Ask DNA Questions in Plain English

For decades, biologists have described the human genome as a book written in a language they can read only letter by letter. A team at Uniklinik RWTH Aachen in Germany has built an AI system designed to translate that book into answers.

What Genolator Does

Genolator is a multimodal AI that takes raw DNA sequences, amino acid chains, and three-dimensional protein structures, then answers plain-language questions about what these molecules actually do.

A scientist can ask, in ordinary English, whether a given protein performs a given function — and receive a reliable, evidence-grounded answer. The system was described in a peer-reviewed study published in Genome Biology on September 24, 2026.

How It Performs

The evaluation results are striking:

Genolator achieves high accuracy in confirming or denying associations between proteins and functions
It outperforms GPT-4.1 on protein function prediction tasks
It also beats smaller domain-specific models that incorporate protein knowledge graphs

This matters because it demonstrates that general AI fluency in English does not automatically translate into biological fluency. Domain-specific architecture and training data still win for specialized scientific tasks.

Why This Is Different from Anthropic's Discovery

The same week Anthropic announced Claude's discovery of the ART enzyme system, Genolator represents a different kind of AI biology breakthrough:

	Anthropic ART Discovery	Genolator
Approach	Autonomous pattern search in DNA databases	Interactive Q&A about molecular function
Output	Novel scientific discovery	Functional annotation of known sequences
Validation	Preprint, not peer-reviewed	Peer-reviewed in Genome Biology
Accessibility	Internal research tool	Published model for broader use

Both advance AI's role in biology, but through different mechanisms. Anthropic's work shows AI can discover what humans missed. Genolator shows AI can explain what humans already have but struggle to interpret.

The Multimodal Advantage

Genolator's key innovation is fusing multiple biological data types — genomic sequences, protein structures, and natural language — into a single model. Previous approaches typically handled one data type at a time, requiring researchers to manually integrate results.

By training on the relationships between sequence, structure, and function simultaneously, Genolator can reason across modalities in ways that single-purpose tools cannot.

Implications for Research and Medicine

The practical applications are substantial:

Drug discovery. Faster functional annotation means researchers can prioritize drug targets with more confidence.

Rare disease diagnosis. Clinicians could query whether a patient's variant affects protein function without waiting for specialized bioinformatics analysis.

Education. Students could explore molecular biology through conversation rather than memorizing lookup tables.

Research acceleration. Tasks that required weeks of manual annotation could be completed in hours.

The Broader Trend

Genolator joins a growing wave of AI tools purpose-built for science:

Anthropic's biology lab and ART discovery
Google DeepMind's AlphaFold and subsequent protein design tools
OpenAI's growing research partnerships in life sciences
Specialized models for chemistry, materials science, and climate modeling

The pattern is consistent: general-purpose LLMs provide a foundation, but domain-specific models deliver the accuracy that science demands.

What Comes Next

The researchers published their full methodology and model details, enabling other labs to build on the work. As genomic sequencing becomes cheaper and more widespread, tools like Genolator could become as standard in biology labs as PCR machines.

The genome is still a book. But for the first time, we have a reader that can discuss what it means.