A LOOK INSIDE: Young Bang, principal deputy assistant secretary for the Army for acquisition, logistics and technology joins Darryl Colvin, joint program executive officer for chemical, biological, radiological and nuclear defense for a walkthrough demonstration of Joint Program Executive Office for CBRND capabilities at Aberdeen Proving Ground, Maryland, in August 2023. (Photo by Matthew Gunther, JPEO-CBRND)
Pacing biological threats with AI using a capabilities-based defense approach.
by Kelly Burkhalter
The national COVID-19 pandemic response emphasized the importance of speed and agility in combatting a biological threat. The pandemic caused millions of deaths, illness and impacted day-to-day life across the globe, emphasizing that despite the type of threat—natural, manmade or otherwise—biological threats can have significant and lasting consequences. DOD historically managed biological threats from a threats-based perspective, that is, creating solutions for specific types of diseases or poisons. What was magnified during the COVID-19 response is that many threats are unpredictable, and the response to unforeseen threats still needs to be swift. This understanding initiated a new approach to biological defense. One that relies on having the best capabilities available to address any threat, rather than solving for a known handful.
A capabilities-based approach invests in platforms to have the infrastructure, processes and skillsets in place and ready to respond to whatever threat may materialize. A platform, like a monoclonal antibody, could be thought of like a foundational starting point that could be easily and quicky adapted to suit different needs. Think of a truck chassis. While there may be different types of cargo containers on the chassis, a new chassis does not need to be developed when it needs to transport different products. The type of containers put on the chassis can change. Artificial intelligence and machine learning (AI/ML) are platform enablers that can support the de-risking, sustainment and preparedness of medical countermeasures (MCM) by modeling the art of the possible before a significant investment is made to further develop an MCM.
The Joint Program Executive Office for Chemical, Biological, Radiological and Nuclear Defense’s (JPEO-CBRND) Joint Project Lead for CBRND Enabling Biotechnologies (JPL CBRND EB) has invested in platform technology and is leveraging AI/ML to support this capabilities-based defense approach. The Generative Unconstrained Intelligent Drug Engineering (GUIDE) program is at the forefront of examining how to use AI/ML to computationally design MCM candidates and test these candidates in dedicated labs. Doing so will accelerate the medical discovery phase and meet the joint force’s needs faster.
HARNESSING THE POWER OF COMPUTING TO STAY PREPARED
The GUIDE gives the medical community a rapid response tool that can increase preparedness posture by speeding up the discovery phase. Using high performance compute powered by capabilities made available to JPEO-CBRND from partners in the Department of Energy, GUIDE reads and analyzes various types of molecular data from studies conducted over the last 50 years and creates algorithms from this data as a prediction tool. AI/ML works exponentially faster than a human and its computations allow it to “see” things on a molecular level that researchers cannot. GUIDE generates upwards of 300 various drug candidate computations that are then tested in a dedicated laboratory by experts and down selected to the best possible candidate. Using this method, GUIDE will identify potential drug candidates that have a high probability of success as measured by meeting four critical attributes (safety, manufacturability, efficacy, pharmacokinetics and pharmacodynamics). Meeting the four critical attributes de-risks the MCM candidate by the medical community by meeting the attributes more accurately, GUIDE provides an effective solution to de-risk an MCM candidate and fast track it into production.
Often, in the MCM development process, most of the cost comes from developing products that cannot be produced or fielded for reasons such as safety, efficacy or manufacturability. By using AI/ML to examine those features early, the MCM that comes out of the discovery phase is more likely to successfully reach production. This de-risking could potentially save millions of dollars. The data from the non-selected candidates is “learned” by GUIDE and able to be drawn upon for future runs.
Recently, GUIDE researchers successfully used AI technology to restore the effectiveness of COVID-19 monoclonal antibodies that have lost function due to viral evolution. Restoring the function of antibodies is a game-changing breakthrough to preserve the life and efficacy of medical countermeasures. For example, at the height of the COVID-19 pandemic, researchers struggled to produce MCMs that were effective against the constant new variants of the virus. Using AI to restore antibodies can help keep up with changing viruses by giving the medical community a chance to prolong the lifespan of MCMs.
“It could take decades and millions of dollars to develop a successful MCM. With GUIDE, AI/ML is accelerating the process by helping us see what a viable MCM option could be in a shorter amount of time,” said Nicole Dorsey, deputy joint project lead for JPL CBRND EB . While this technology gives the medical community a head start, it does not replace the human researcher. It would, instead, make the human researcher’s job more important.
GUIDE THE WAY: GUIDE is an interagency program between JPL CBRND EB, the Department of Energy and other interagency, academic and industry partners. GUIDE’s mission is to use its integrated computational and experimental capabilities to accelerate drug development for the warfighter by harnessing the power of advanced simulation and machine learning. (Graphic by Maya Munk, JPEO-CBRND)
AI/ML AS A TOOL AND PARTNER, NOT A REPLACEMENT
Even if AI/ML can generate a world of possibilities, human researchers are still needed to test and select the best fit, understanding the broader context of the needs and requirements, and applying scrutiny and methods to eliminate the possibility of data bias or other nuances that computers are unable to discern.
Since GUIDE is an agnostic system that analyzes data from a molecular level, there is no potential gender or racial bias to consider. Once the selected computations enter a clinical trial, the medical community can ensure that bias is avoided by enrolling diverse human subjects in the studies to understand the medical impact across populations. Computations can only produce data; humans must set the conditions to ensure the final products minimize and account for any discrepancies across human subjects.
GUIDE uses both historical data and purpose-built datasets which do not have these inherent biases. Rigor in testing should account for diverse populations and circumstances to understand how drug candidates engage and interact with human subjects. Humans still need to have a role in programs that use AI/ML technology to provide that discerning eye and ensure there are safeguards to minimize bias. While a machine can read the data, scientists must interpret the data, apply it and figure out if what the data suggests can be possible.
“AI is great, but it can only see so much, especially when we are looking at molecules,” said David Bailey, acting director for advanced technology platforms at JPL CBRND EB. “We can see some ‘what-if’ molecules, but we need the experts in the medical community and the clinical trial process to show the efficacy and safety is validated across various subject sets.”
GETTING TO THE FIELD, FASTER
Response time is critical to defeat any threat. AI/ML can help programs like GUIDE get a head start to finding solutions, but speed and agility must be embraced throughout the entire process, including procedures and collaboration, to prevent stalls in fielding a solution. AI/ML can speed up potential solutions, but the next steps need to move just as quickly to defeat an evolving biological threat.
“GUIDE gives the community best potential starting molecule to fight a threat, but that molecule then needs to be further developed. This rigorous testing and evaluation can only happen from the support of the U.S. government, industry and academic partners to get it over the finish line,” said Dorsey.
The DOD must work together to apply and validate findings from AI/ML and use all the resources at its disposal to move a candidate from discovery to Phase I clinical trials and then on to production. Some of those resources could include using opportunities such as Expanded Access Protocol to allow for emergency access to MCMs that are currently under development, working with the whole of government to develop requirements rapidly or investigating how tools like AI/ML could be used in the post-discovery phases of the MCM development process.
“We are using the power of technology to pull data that the Army, U.S. government and industry collected over decades and reuse it,” Dorsey said. “Next, we optimize the data and build it into a data model to address real world threats for industry and government partners. This allows us to send the joint force what they need most in the field, providing a defense against whatever threats they may encounter.”
CONCLUSION
The best way to react to the unknown is to be prepared. AI/ML is one tool that can support preparedness by accelerating MCM discovery. Tools like these allow DOD to hit the ground running in the face of a threat by saving time and money. To get these MCMs from concept to production requires human coordination and partnership within the community to ensure that the potential candidates generated by AI/ML and experimentation can be operationally relevant and accomplish its mission of saving lives. Realizing the full benefits of technology requires collaboration, responsible use and scientific rigor.
“AI/ML is just one of the tools in our capabilities-focused approach to biodefense,” said Bruce Goodwin, joint project lead for CBRND EB. “Technology like this allows us to carry out our mission to enable joint force resiliency, provide rapid response in the face of a threat and provide operationally relevant solutions that can give our joint force the support they need at the time they need it.” Fiscal year 2024 marks the second year of GUIDE as a program and, with the support of its partners across the government, it achieves new levels of preparedness and makes the unknown manageable for the joint force.
For more information, contact the JPEO-CBRND Public Affairs Office at usarmy.apg.dod-jpeo-cbrnd.mbx.jpeo-cbd-public-affairs-office@army.mil or visit https://www.jpeocbrnd.osd.mil.
KELLY BURKHALTER is a lead associate at Booz Allen Hamilton where she supports strategic communications programs in support of U.S. Army and DOD organizations such as JPEO-CBRND. She holds an M.A. in communications from Johns Hopkins University and a B.A. in English and journalism from Syracuse University.
