Lookout | When synthetic biology encounters AI biosecurity risks, vigilance is required

Experts point out that artificial intelligence technology helps synthetic biology overcome a fundamental challenge, that is, to find the optimal solution from a large number of complex systems

The biosafety risks posed by non-professional groups entering synthetic biology research with the help of AI may not be affiliated with any research institution or organization and are excluded from the review and management of the scientific community and institutions

Malicious actors may spoof AI protections by spoofing and using ChatGPT to obtain information such as viral sequences with pandemic potential, technologies to assemble viral genomes, necessary lab supplies, and companies that can provide these devices, all of which could make known viruses more dangerous

"Outlook" Newsweek reporter Zhang Jianxin Li Yating

Anti-cancer drugs, renewable biofuels, meatless burgers that taste like meat, the scent of extinct flowers... In recent years, the booming synthetic biology is quietly changing people's lives.

Synthetic biology is an interdisciplinary field that combines biology, engineering, and computer science to design and build biological systems with new capabilities. "Synthetic biology is like the process by which cows produce milk." Some insiders have explained the research content of synthetic biology as follows, "In synthetic biology, chassis cells are equivalent to cows, and various raw materials are equivalent to low-value forage. Researchers modify chassis cells through gene editing technology, add low-cost raw materials and then ferment a series of high-value chemical or biological raw materials, which is equivalent to producing milk."

The "milk" produced by synthetic biology has a wide range of application scenarios in many fields such as medical and health, chemical industry, agriculture, and food. McKinsey & Company has released data that estimates that the economic impact of synthetic biology and biomanufacturing will reach $100 billion by 2025; In 2030~2040, the annual economic impact of synthetic biology will reach 1.8 trillion~3.6 trillion US dollars.

With the rapid development of artificial intelligence (AI) technology, more and more researchers have begun to explore the application of AI to synthetic biology, which has further promoted the progress of synthetic biology in chemical manufacturing, novel materials, human health and environmental protection.

However, the deep integration of artificial intelligence and synthetic biology also brings more risks to a certain extent. In an interview with the "Outlook" newsweek reporter, Wang Fangzhong, associate professor of the Biosecurity Strategic Research Center of Tianjin University, said that artificial intelligence may induce more biosecurity risks, such as increasing the probability of biosecurity risks and increasing the degree of biosecurity risks.

Experts pointed out that reducing biosecurity risks requires efforts to strengthen the strategic deployment of biological defense, attach importance to talent training in related fields, and improve the supervision of synthetic products.

Artificial intelligence empowers the development of synthetic biology

The purpose of synthetic biology is to design biological systems that meet standards, and the "design-build-test-learn" cycle (DBTL) that uses engineering predictability to control the construction of complex systems based on engineering design principles has gradually become the core strategy of synthetic biology. In biomanufacturing, the DBLT cycle of four stages can successfully build the required cells and produce the right product.

With the continuous improvement of people's demand for "milk", artificial intelligence technology has been introduced and applied to the field of synthetic biology research. At present, artificial intelligence has been widely used in synthetic biology fields such as gene circuits, metabolic engineering, and genome engineering, which not only helps to improve the research and development efficiency of bioengineering, but also lowers the research threshold and becomes a powerful tool to promote the development of synthetic biology.

-- Improving the efficiency of bioengineering research.

On August 9, Clinical Pharmacology and Therapeutics, a leading journal in the field of experimental and clinical medicine, published a study introducing inClinico, an artificial intelligence clinical trial prediction engine, and its accurate prediction of phase II. to phase III transformation results of multiple clinical trials.

"Currently, more than half of phase II clinical trials have failed, resulting in hundreds of billions of dollars and years of efforts being wasted. There are many complex reasons behind the failure of clinical trials, and AI has a unique advantage in solving this problem. The authors believe that AI tools to accurately predict the transformation success rate of clinical trials Phase II to III can help researchers guide successful clinical trials earlier in the drug discovery process.

Experts point out that artificial intelligence technology helps synthetic biology overcome a fundamental challenge, that is, to find the optimal solution from a large number of complex systems. Previously, the cell engineering goal of synthetic biology (i.e., reverse design) could only be achieved through a lot of trial and error, which was inefficient due to the inability to predict the outcome of bioengineering, while artificial intelligence could use open data and experimental data to predict the outcome.

Not only in data analysis and prediction, artificial intelligence can also help improve work efficiency, reduce costs, and shorten the R&D cycle in genome design and optimization of synthetic organisms, protein engineering and design, and biological system simulation and optimization.

For example, AI can help researchers process and analyze large-scale biological data, including information such as genomes, proteomes, and metabolomes. Using machine learning and data mining, AI can spot patterns and trends in vast amounts of information to predict the behavior and response of biological systems.

"At present, scientists have built artificial intelligence models through small-scale test results or using viral gene sequences stored in databases to predict whether the virus will escape detection after mutation without destroying other characteristics of the virus." Wang Fangzhong introduced that the study has established relevant models for avian influenza virus and HIV virus. As more viral sequences are discovered and published, more predictive mutation models for different species of viruses are expected to emerge.

-- Lowering the threshold for synthetic biology research.

Recently, a netizen posted his conversation with ChatGPT about synthetic biology. The content of the dialogue shows that ChatGPT can not only answer relatively popular science questions such as "how to explain synthetic biology to an eight-year-old child" and "how young people start to engage in synthetic biology", but also answer professional questions such as "designing a gene sequence that can express GFP in high concentrations in E. coli", so that the difficulty of learners to learn synthetic biology-related professional knowledge is greatly reduced.

At the same time, the continuous commercialization of synthetic biotechnology services based on artificial intelligence technology has further reduced the difficulty of synthesizing related products. For example, there are already biotechnology companies that can provide customers with overall design solutions for synthesizing new substances, using artificial intelligence and biocomputing technology for virtual screening and large-scale mass production. With the help of artificial intelligence, synthetic biology technology has become more widely accessible.

Industry insiders pointed out that with the maturity of low-cost and simple and practical experimental tools, the popularization of biotechnology knowledge, the open source of synthetic biology research and development, and the establishment of new biological design-construction institutions, the field of synthetic biology research is not only the territory of highly skilled senior experts with rich biomedical knowledge, but also will attract a large number of scientists or amateurs with no biological experience to participate.

Induce more biosecurity risks

In 2017, a Canadian virologist spent only $100,000 to synthesize horsepox virus, a "close relative" of smallpox, raising concerns about whether smallpox would make a comeback.

While artificial intelligence promotes the rapid development of synthetic biology, it may also lead to an increase in the probability of biosecurity risks and an increase in the degree of harm, bringing more challenges to existing biosafety and biosecurity governance.

Notably, biosecurity risks posed by the intentional release of biologics or materials by malicious actors may also be further amplified with the assistance of artificial intelligence.

-- Increasing the probability of biosecurity risks.

Industry insiders believe that the combination of artificial intelligence and synthetic biology not only lowers the threshold of research, but also increases the probability of synthetic organisms being used maliciously.

"Malicious actors only need basic code running capabilities to turn a useful medical tool into a generator of potentially lethal molecules." Wang said that malicious actors can use open source model code and public databases to create new toxin databases.

In the investigation, the reporter found that in open academic journals and websites, there have been a large number of schemes on how to use artificial intelligence technology to optimize the genetic design of chemical synthesis pathways. For example, an article describes a method that uses logic to model specific toxic molecules by inverting machine learning models to drive the model to create 40,000 new toxin molecules in 6 hours.

"With the help of artificial intelligence, the time and economic cost of synthesizing new toxins will only be lower, and the biosecurity risks caused by them will be more hidden." Wang Fangzhong said.

In addition, Xue Yang, associate professor at the Biosafety Strategic Research Center of Tianjin University, believes that biosecurity risks caused by non-professional groups entering the field of synthetic biology research with the help of artificial intelligence may not belong to any research institutions or organizations and are excluded from the review and management of scientific communities and institutions, so the biosecurity risks caused by them will be more difficult to monitor and manage.

-- Increasing the degree of harm to biosecurity risks.

Enhancing the immune evasion ability of the virus is an important means of enhancing the virulence of the virus, which will become less difficult with the help of artificial intelligence.

Experts told reporters that the synthesis of some toxins may bring risks such as increased virulence of pathogens or pathogenic pathogens, improved transmission capacity of pathogens, changes in pathogen hosts, ineffective diagnostic measures, ineffective vaccines, and antibiotic resistance, and the emergence of these conditions may have a fatal impact on the human body.

For example, malicious actors may use synthetic biology techniques to specifically edit microbes in the gut flora that affect immune function, spreading them through air, water, or food, weakening human immunity. Once such practices are engineered by malicious actors, they may pose a risk of large-scale virus infection.

"At present, our ability to identify and attribute attacks is weak, and the mitigation measures are inadequate, and even the slightest malicious manipulation of the gut microbiota can gradually change a person's physiology and behavior." Zhang Weiwen, director of the Biosecurity Strategic Research Center of Tianjin University, said.

In addition, AI chatbots can also be exploited maliciously as tools to enhance the virulence of known viruses.

In an MIT classroom test, ChatGPT came up with 4 potential pandemic pathogens in an hour and helped students identify which ones were most lethal, providing a list of companies that might assist with DNA synthesis, and even telling students how to trick these companies into getting services.

"Malicious actors may spoof AI protections by spoofing and using ChatGPT to obtain information such as viral sequences with pandemic potential, technologies to assemble viral genomes, necessary lab supplies, and companies that can provide these devices, all of which could make known viruses more dangerous." Wang Fangzhong said.

Reducing biosecurity risks requires multiple efforts

A number of experts interviewed pointed out that although the deep integration of synthetic biology and artificial intelligence may have certain biosecurity risks, it should not be "choked". The government, academia, and industry should work together to formulate countermeasures through a combination of "cautious prevention" and "product orientation" to ensure technological progress while minimizing risks.

Strengthen the strategic deployment of biological defense. Wang Fangzhong suggested that relevant government departments should strengthen the strategic deployment of biological defense, research and formulate long-term development plans, attach importance to research in related fields, and increase policy and financial support. Promote the establishment of national high-end think tanks for biosecurity, and form a gathering place for high-end biosecurity talents.

In addition, experts suggest that academia should strive to enhance practitioners' awareness of risk and responsibility, and include relevant safety training in the project application, approval and implementation stages. Practitioners should maximize their ability to identify, assess, respond and mitigate risks.

Pay attention to the cultivation of talents in related fields. Zhang Weiwen said that attention should be paid to the cultivation and education of talents in related fields, and more government, academia and industry personnel with relevant professional knowledge and experience should be absorbed into the existing synthetic biology risk management system, and the risk review and consulting expert database should be improved. At the same time, we should focus on cultivating the next generation of artificial intelligence and synthetic biology risk management professionals with an international perspective to achieve continuous updating and iteration of talents.

Improve the supervision of synthetic products. Xue Yang suggested that the application of artificial intelligence technology in the field of synthetic biology has further blurred the boundary between chemical and biological weapons. Relevant people said that the establishment of a risk reporting mechanism arising from the deep integration of artificial intelligence and synthetic biology should be encouraged, and a "white list" and "black list" of chemical synthesis or biosynthesis should be established, and companies engaged in chemical or biological synthesis and high-throughput screening companies should decide whether to provide services based on the content of the list.

In addition, the industry should promote the commercialization of synthetic products in a differentiated manner, promote the introduction of relevant laws and regulations as soon as possible, and clarify the declaration, approval and market access standards for various products. ■