Track trials, findings, and clinical evidence as it lands.
Search papers, labs, and topics across Lattice.
This paper introduces a position-independent watermarking technique for synthetic DNA generated by genomic language models, enabling provenance tracking without prior knowledge of sequence origin or tokenization. The SynthID tournament watermark was embedded in two models, Carbon and GENERator-v2, and a robust verifier demonstrated perfect detection of marked sequences across various mutations and search parameters. This approach offers a statistically sound method for identifying model-generated DNA, with minimal impact on model performance and high resilience to random edits.
Model-generated DNA can now be traced with perfect accuracy, even after single-base edits, using a watermark detectable without knowing its precise location or strand.
Genomic language models can write synthetic DNA that carries no record of its origin. A generation-time watermark could provide a provenance signal, if a verifier can detect it using only the DNA sequence, a key, and published detector settings, without knowing where the generated region starts, which strand it lies on, or how it was divided into six-base tokens. We embed the SynthID tournament watermark in two genomic language models, Carbon and GENERator-v2. The verifier searches both strands, every start position, and four window lengths, and corrects its decision for the whole search. In a detection cohort of 1,544 held-out prompts, it found all 3,088 marked sequences per model, unedited and after one substituted, inserted, or deleted base. For ordinary sequences, the one-sided 95% upper confidence bound on the false-positive rate was at most 0.850%; one bound for the wrong-key control reached 1.015% (Carbon, after a deletion). In a development cohort, neither model showed a detectable change in likelihood or in predefined sequence measures. Under random edits placed without reference to the detector, detection stayed complete or nearly complete up to a 2% per-base edit rate. These results show statistical detection, not biological function, secret-key security, or robustness to an editor who sees the detector.