SynthID Bio: A New Tool for Verifying AI-Generated Biological Sequences and Structures

A new tool has been developed to address the growing concern of biosecurity in synthetic biology. SynthID Bio, a family of watermarking methods, is designed to strengthen biosecurity and scientific integrity by embedding an imperceptible signature directly into biological code. This approach ensures that watermarks are verifiable not just on digital models but also on synthesized physical proteins themselves – all while preserving their biological function in laboratory testing.

SynthID Bio builds upon the success of SynthID, a tried-and-tested watermarking tool developed by Google DeepMind. The new method adapts its approach depending on the type of data, subtly guiding the choice of amino acids for sequences and adjusting atomic coordinates for predicted 3D structures to create a reliable signal for detection.

In experiments, these adjustments did not compromise the protein’s biological function, which is essential for effectively treating disease and advancing scientific research. The researchers verified their approach by using AlphaProteo, a binder design method, alongside a SynthID Bio-enabled version of ProteinMPNN, a commonly used protein sequence generation method.

The team tested watermarked designs on three target proteins – VEGF-A, the SARS-CoV-2 spike protein RBD, and PD-L1. The results showed that watermarked designs matched the hit rate, binding affinity, and natural sequence diversity of unwatermarked versions, successfully creating the first-ever watermarked and biologically functional protein binders.

SynthID Bio also fine-tunes a small part of AlphaFold 3’s diffusion network to build the ability to watermark directly into the model’s weights. This ensures that predicted 3D coordinates inherently carry a detectable signature regardless of who runs the model, preserving AlphaFold 3 prediction accuracy while offering near-perfect detectability.

Biosecurity relies on layered defenses – think of it like a ‘Swiss cheese’ defense model, where multiple independent safety measures work in tandem to cover each other’s blind spots. SynthID Bio serves as an important verification layer embedded in the biological design itself, linking designs to the model developer and empowering developers to lead on safety.

SynthID Bio can also be paired with provenance metadata approaches or central repositories of AI-generated biological data to better identify and track AI-generated proteins. The team is researching how to apply watermarking to more complex biological objects, such as genomes, in ongoing work with the Hie lab at Stanford University and Arc Institute.

The extension of this work to watermark DNA of bacteriophages is a collaboration between Google DeepMind and the Hie lab at Stanford and Arc Institute. Key contributions were made by Jeremy Ratcliff, Aleks Petrov, Alexander I. Cowen-Rivers, David Stutz, Elisa L. H. Wong, Victor Martin Palacios, Francesca Pietra, Alfred Piccioni, Tristan Oliver Kwan, Tor Lattimore, Sumanth Dathathri, and Pushmeet Kohli from Google DeepMind.

SynthID Bio has the potential to address some of the biosecurity risks associated with genome design. The team is committed to responsible innovation and open-sourcing the code and in vitro data, releasing the weights to the research community so they can build on this work.

The development of SynthID Bio was initiated by Pushmeet Kohli, led by Alexander I. Cowen-Rivers and David Stutz, with key engineering and research contributions from Guillermo Ortiz-Jimenez, Jeremy Ratcliff, Vinicius Zambaldi, Lindsay Willmore, Josh Abramson, Harshnira Patani, Christina Kouridi, Florian Stimberg, Mel Vecerik, Alex Chu, Sukhdeep Singh, Sumanth Dathathri, Eliseo Papa, Valentin De Bortoli, Arnaud Doucet, Jue Wang, and Sven Gowal.

SynthID Bio can help maintain the integrity of databases such as the Protein Data Bank, UniProt, and GenBank. These databases play a vital role in scientific advancements but mislabeled entries can have an outsized negative impact in biosecurity decision-making.

The team is also researching how to apply watermarking to more complex biological objects, such as genomes, in ongoing work with the Hie lab at Stanford University and Arc Institute. This collaboration has the potential to address some of the biosecurity risks associated with genome design.

SynthID Bio can provide an automated verification signal, proving that an order originated from a trusted model with built-in safeguards. This is especially vital for DNA synthesis screening, which sits on the frontlines of biosecurity. Converting digital protein designs into physical molecules requires placing an order with DNA synthesis providers who screen requests against databases of known threats.

SynthID Bio can help streamline screening by providing an automated verification signal that proves an order originated from a trusted model with built-in safeguards. This is especially vital for DNA synthesis companies, which have an important role in helping AI-driven innovation scale responsibly.

The development of SynthID Bio has the potential to address some of the biosecurity risks associated with genome design and synthetic biology. The team is committed to responsible innovation and open-sourcing the code and in vitro data, releasing the weights to the research community so they can build on this work.