TanzaniaGoogle DeepMind Watermarks AI Proteins
Google DeepMind has developed SynthID Bio, a technology that watermarks AI-designed proteins while preserving their biological function, with potential applications in biosecurity and open scientific databases. The method has been tested in laboratory settings with positive results.
Global Newsflix 2 min read
Google DeepMind researchers have successfully watermarked AI-designed proteins, a breakthrough that could help track the origin of synthetic biological designs. The new method, SynthID Bio, embeds a detectable signature into a protein's amino acid sequence or its predicted three-dimensional structure.
According to Pushmeet Kohli, Chief Scientist at Google Cloud and VP Science at Google DeepMind, the result is the 'successful synthesis of AI-designed proteins that are both functional and watermarked'. This achievement is significant because altering a protein can change its function, making it challenging to watermark without affecting its performance.
SynthID Bio uses two approaches: one for protein sequences and another for three-dimensional protein structures. For sequences, the system subtly changes how amino acids are selected while an AI model generates the sequence. For structures, the researchers fine-tuned part of AlphaFold 3 to build the watermark into the predicted atomic coordinates.
Laboratory tests have shown that watermarked protein binders remain functional across three targets: VEGF-A, PD-L1, and the receptor-binding domain of the SARS-CoV-2 spike protein. The study found no significant population-level differences in binding affinity between watermarked and non-watermarked binders.
The detection rate for the structural method achieved above 99.8% at a 0.1% false-positive rate in the researchers' tests, without reducing key structural accuracy measures. However, the results were not identical across every measure, indicating that SynthID Bio is a proof of concept rather than a finished biological provenance system.
The development of SynthID Bio has potential implications for biosecurity and the integrity of open scientific databases. By embedding a verifiable watermark into biological designs, researchers can track the origin of AI-designed proteins and prevent potential misuse.
Google DeepMind is working with Stanford University and Arc Institute to further develop SynthID Bio. The technology is expected to strengthen biosecurity and preserve the integrity of open scientific databases.
Key facts
- Google DeepMind has developed SynthID Bio, a technology that watermarks AI-designed proteins.
- The method embeds a detectable signature into a protein's amino acid sequence or its predicted three-dimensional structure.
- Laboratory tests have shown that watermarked protein binders remain functional across three targets.
- The detection rate for the structural method achieved above 99.8% at a 0.1% false-positive rate.
Three perspectives
Neutral
The development of SynthID Bio is a significant breakthrough in the field of synthetic biology, with potential applications in biosecurity and open scientific databases. However, the technology is still in its early stages, and further research is needed to fully understand its implications. As the technology continues to evolve, it will be important to monitor its development and potential impact.
Positive
The successful development of SynthID Bio demonstrates the potential of AI in advancing synthetic biology and improving biosecurity. The technology has the potential to prevent potential misuse of AI-designed proteins and promote transparency in scientific research. With further development, SynthID Bio could become a valuable tool in the fight against bioterrorism and the promotion of open scientific collaboration.
Negative
The development of SynthID Bio raises concerns about the potential risks and unintended consequences of watermarking AI-designed proteins. The technology could be used to track and control the use of synthetic biological designs, potentially stifling innovation and limiting access to important research. Additionally, the use of watermarks could create new vulnerabilities in biological systems, potentially leading to unforeseen consequences.
Your week in news, every Wednesday
The biggest stories of the week for your edition, in your inbox every Wednesday morning, your time. Free, and you can unsubscribe at any time.