Understanding the Detection of Bioengineered Viruses
Written on
Chapter 1: The Emergence of Speculation
From the moment the coronavirus made headlines, theories emerged suggesting it might have been deliberately created, possibly stemming from experiments at one of several laboratories in Wuhan. Some political figures promoted the narrative that the virus originated from a scientific facility, prompting the White House to urge intelligence agencies to investigate potential lab connections.
Most researchers concur that, based on genetic evidence, the virus likely transferred from animals to humans. On April 30, the U.S. Office of the Director of National Intelligence stated, representing the 17 agencies of the U.S. intelligence community, that "the Covid-19 virus was not manmade or genetically modified." They decided to continue probing two primary possibilities: the more plausible theory that the virus transitioned from animals to humans, and the less likely scenario of a naturally occurring virus accidentally released from a lab.
Thus, the U.S. intelligence community "aligns with the broad scientific consensus," as the statement noted, that the virus was not human-made. But how did they reach this conclusion? While the full extent of their investigation remains undisclosed, one program known as FELIX specifically explored this hypothesis. The analysis conducted by FELIX indicated that the virus was not engineered using "foreign" genetic sequences, confirming that SARS-CoV-2, responsible for Covid-19, was not artificially created or modified.
Detecting signs of bioengineering is inherently challenging for any organism. Various methods can reveal whether a virus was engineered, and conversely, there are numerous approaches to engineering a virus, leading to a persistent struggle filled with uncertainty.
Section 1.1: The Role of FELIX
FELIX, which stands for Finding Engineering-Linked Indicators, is managed by IARPA (Intelligence Advanced Research Projects Activity). IARPA engages in high-risk research and the development of next-generation technologies under the Office of the Director of National Intelligence. In 2018, FELIX began financing six external teams to create tools capable of identifying indicators of bioengineering. These genetic markers serve as clear signs of manipulation within an organism's genome.
A genome encompasses the complete set of genetic bases constituting an organism. In DNA, these bases include A, G, C, and T, while in RNA, they consist of A, G, C, and U. Arranged in sequences, these bases describe an organism's genetic makeup.
According to FELIX program manager David Markowitz, PhD, engineering fingerprints within a genome can manifest in several ways. These may include foreign genetic material in a sequence or irregular duplications, insertions, or deletions of bases. Other indicators, as noted by Isaac Plant, PhD, a former FELIX graduate student, can feature sequences linked to antibiotic resistance and short sequences called "scars," which indicate alterations to a DNA sequence. While there isn't an exhaustive catalog of "engineering sequences," databases like Addgene provide molecular tools for DNA manipulation.
FELIX's instruments can ascertain whether someone has appropriated biological intellectual property — for instance, if a specialized yeast strain from one company appears in a competitor's laboratory — and assess the natural occurrence of new pathogens. SARS-CoV-2 presented FELIX with its first significant real-world challenge.
"We spent 18 months developing the initial working prototypes of our engineering detection platforms," Markowitz explains. "They were ready to tackle SARS-CoV-2 when this biosecurity threat emerged."
In January, a team from the MIT-Broad Foundry utilized FELIX tools to "validate online claims suggesting that SARS-CoV-2 was produced in a laboratory," as stated on IARPA's website. Although the results of that evaluation are not entirely public, a notice on the page indicates that the system compared the virus's genome to 58 million known genetic sequences, encompassing both closely and distantly related viruses. Within 10 minutes, the tool determined that the virus's genetic structure corresponded more closely to naturally occurring coronaviruses than any other organisms: "This analysis suggests that no foreign species sequences were engineered into SARS-CoV-2," stated IARPA.
However, as Filippa Lentzos, PhD, a senior research fellow at King's College London specializing in biosecurity, cautions, this finding does not eliminate the possibility of engineering; it simply means that the virus was not assembled using specific methods.
Section 1.2: The Challenges of Detection
The MIT-Broad Foundry team that investigated SARS-CoV-2 declined to participate in an interview. While limited information is available regarding their methodologies, other FELIX-funded teams have been more forthcoming.
Eric Young, PhD, who examines yeast engineering at Worcester Polytechnic Institute, pivoted his focus towards biosecurity after observing concerns expressed by policymakers and ethicists at synthetic biology conferences about the implications of easily creating custom organisms.
"What we're developing could significantly benefit civilization," Young states, highlighting advancements in pharmaceuticals. "However, history shows that many technological breakthroughs have been repurposed for weaponization." So far, he notes, there are no known instances of synthetic biology being employed to create a bioengineered weapon.
The tool Young developed for FELIX first sequences an organism's entire genome. Then, it returns the genetic data to researchers with annotations identifying any manipulated genetic components. It assesses engineering by comparing the sequence to a standard list of yeast engineering sequences or a custom list provided by the user. "Future versions will utilize machine learning to identify engineered DNA sequences without needing a predefined list of components," Young explains. Currently, it operates specifically with yeast sequences.
Ginkgo Bioworks, another recipient of FELIX funding, approaches the detection challenge through computational biology and its own database of recognized engineered sequences. The company often operates on the other side of the equation, designing and engineering organisms itself. "For this initiative alone, we produced over 6 million simulated engineered genomes reflecting various genetic engineering techniques and design styles, selected by IARPA and national laboratories," says Joshua Dunn, PhD, the company’s head of design.
"We aimed to be the ethical hackers in this context," Dunn elaborates. To achieve this, Ginkgo employs multiple methods. First, they compare a sequence to known natural reference sequences, similar to the MIT-Broad Foundry's approach with SARS-CoV-2. Then, they search for recognized engineering signatures within the same input sequence. A third method analyzes the distribution of the genetic alphabet. Finally, all gathered data feeds into a fusion engine the team is developing that evaluates the three results collectively to reach a final conclusion.
At the Draper Laboratory, another FELIX group led by Kirsty McFarland, PhD, is concentrating on detecting engineered organisms in environmental samples teeming with life, such as soil or water. They are developing distinct methods aimed at identifying two forms of bioengineering: unknown changes to known organisms and known or suspected alterations to potentially unknown organisms. SARS-CoV-2, as a novel pathogen, falls into the latter category.
The first method compares identified genomes to a database of anticipated gene sequences. The second seeks sequences of interest — potential engineering signatures — within the genomes of previously unknown organisms. Currently, the second method can detect a single engineered organism amidst a million or more natural entities.
At Harvard, a team led by Elizabeth Libby, PhD, is creating a "biosensor": an engineered cell designed to detect other engineered organisms. Naturally, the Boston team began sketching the sensor's concept on a Dunkin' Donuts napkin. "It can essentially absorb any nearby DNA," Libby explains. Then, it identifies whatever DNA signatures they have preprogrammed, amplifies the signal, and lights up to indicate a match.
For FELIX, they have programmed the biosensor to respond to common engineering signatures. However, the technology has broader applications. "We envision that in the future, if you wanted to identify a pathogen in, say, an air-handling system or on a hospital surface, you could have a disposable cartridge that continuously senses," Libby states. If it detects "the thing of interest" — whether an engineering indicator, a coronavirus, or another genetic sequence — the cartridge illuminates and sends an alert. This technology could be invaluable for identifying potential threats in settings like an office during a pandemic.
Chapter 2: The Limitations of Current Methods
Currently, these methods share a common limitation: they depend on records of known organisms or recognized engineering signatures. In essence, they require a reference catalog for comparison. The statement on IARPA's website suggests that the MIT-Broad Foundry's approach also relies on this principle.
"You will never achieve complete certainty in detecting engineered organisms," cautions Plant. The challenge lies in the fact that most analyses depend on existing data and assumptions about the diverse nature of engineering. "Those attempting engineering are just as knowledgeable as those trying to detect it," Plant remarks, suggesting a continuous arms race. "You will never fully eliminate the ability to detect engineered organisms."
The MIT-Broad Foundry's SARS-CoV-2 analysis dismissed the notion that the virus was constructed from parts of other organisms — a prevalent method among FELIX teams. However, numerous pathogens possess genomes that are not included in databases. "You cannot account for all viruses that have been discovered but not shared," notes Alina Chan, PhD, a postdoctoral researcher at the Broad Institute. Many researchers take years to publish their findings. "I could withhold data for 10 years if I chose to. There is no regulation against that."
While such delays are commonplace, they hinder initiatives like FELIX, as some engineering detection tools function like plagiarism checkers, examining whether an organism harbors appropriated genetic information. A careless plagiarist might simply copy and paste entire passages, making them easy to spot. A more sophisticated detector can recognize when someone has synonymized a few words, while an advanced detector might discern if a term paper is modeled after a published article. All detection methods, like engineering detection tools, require a comprehensive catalog of published materials. If a previously unknown virus is manipulated in the wild, it becomes significantly more challenging to identify.
Plant offers a different analogy for the challenge of detecting engineered organisms. "Identifying whether an organism has been artificially created is akin to determining if a new word has been intentionally invented," he explains. "Achieving that with absolute certainty requires knowledge of all existing words as well as those being unintentionally created. That, much like detecting engineered organisms with complete accuracy, is unfeasible." Programs like FELIX may never fully resolve whether an organism has undergone engineering.
FELIX's existing analysis of SARS-CoV-2 would recognize instances where scientists integrated publicly known sequences into pathogens. A more astute bioengineer would likely employ more discreet techniques, as Chan points out: "If you wanted to infiltrate an event unnoticed, you would disguise yourself as someone unknown, rather than imitating multiple well-known celebrities."
What FELIX has established is evidence that SARS-CoV-2 does not consist of disparate parts from various organisms. This hypothesis is valuable to rule out. "This was a limited application of a toolkit," states Gregory Koblentz, PhD, associate professor and director of the graduate biodefense program at George Mason University.
Even if FELIX were to refine its genetic engineering detection capabilities, its findings would likely generate more questions than answers. In the event of an outbreak, intelligence and health officials would still seek to understand where the pathogen was engineered, who was responsible, and the motives behind it. "That requires significantly more information," Koblentz emphasizes — not just in technical terms but also regarding intelligence and law enforcement.
Although FELIX does not address these inquiries, it raises alternative questions. While its primary goal is to enhance biosecurity, its technology carries dual-use implications: it can be used for offensive purposes as well. If one understands how to detect bioengineering, they can potentially conceal their own activities. In this regard, FELIX may serve as a subtle display of capability to the international community, indicating offensive potential without breaching bioweapons conventions. IARPA did not respond to inquiries about dual-use implications before publication.
Programs like FELIX convey an additional message globally. "[This research] is motivated by the belief that the spread of increasingly advanced biotechnology is creating new threats that we are ill-equipped to detect," Koblentz explains. "This is an effort to avert another Pearl Harbor, to prevent a surprise attack. FELIX is just one manifestation of that."
IARPA has other initiatives; DARPA has its own as well.
In essence, these programs signal to the world that the United States perceives biothreats as significant risks. This perception could, as Lentzos suggests, lead other nations to pursue more pathogenic research to avoid falling behind. "In attempting to safeguard yourself," she warns, "you may inadvertently create a greater threat."