The Defense Threat Reduction Agency’s Chemical and Biological Technologies Department (DTRA CB) is leveraging two hallmarks of modern computing — artificial intelligence (AI) and machine learning (ML) — to rapidly identify potential medical targets that could be further developed into measures that counter chemical and biological threat agents. AI is broadly defined as “the science and engineering of making intelligent machines, especially intelligent computer programs,”1 and ML is the use of “methods from neural networks, statistics, operations research and physics to find hidden insights in data without being explicitly programmed where to look or what to conclude,”2 DTRA CB seeks to apply AI and ML technologies to reduce traditional drug-discovery timeline from years to months.
To thoroughly understand the challenges involved in undertaking this drug-discovery revolution, DTRA CB hosted a two-day, interagency workshop that included 90 scientists. The workshop adjourned with an action plan for DTRA CB: to the extent possible, standardize data collection and coordinate research assumptions across projects to take advantage of ML and AI technologies for identifying medical countermeasures for chemical and biological threat agents.
The intent of the workshop on AI and ML aligned, overall, with DoD’s commitment to build a more lethal force, which is one of three Lines of Effort presented in the 2018 National Defense Strategy. DoD wants its armed forces to possess “decisive advantages for any likely conflict, while remaining proficient across the entire spectrum of conflict.”3 To maintain force lethality during a mission, warfighters must have access to medical countermeasures when they encounter a known or emerging chemical and biological threat agent. Without this access, the agent will harm warfighters and reduce their ability to efficiently complete their mission. DoD recommends various strategies to build a more lethal force, including modernization of DoD’s key capabilities and investing in AI and ML “to gain competitive military advantages,”4 Maintaining technological superiority in AI and ML has strategic, national security implications for the U.S. government because near-peer adversaries are investing significant resources to become leaders in the two fields.
After months of planning, DTRA CB’s workshop on AI and ML occurred in October 2019. Participants represented various federal entities: DoD agencies and service laboratories, Department of Energy national laboratories, federally funded research and development centers, and Department of Health and Human Services. The workshop had three aims:
- Identify existing, and developmental, AI and ML capabilities relevant to drug discovery.
- Define technical gaps and investment opportunities to improve state-of-the-art AI and ML platforms.
- Build a road map for DTRA CB to use AI and ML applications.
Workshop participants learned about federal efforts in applying AI and ML technologies for drug-discovery purposes. A panel of participants shared funding priorities related to AI and ML. In two facilitated discussions, participants shared ideas on lessons learned from using AI and ML technologies. One facilitated session focused broadly on issues related to integrating, assimilating, and storing large amounts of data produced by DoD and national laboratories. The other facilitated session focused on issues related to applying AI and ML in the discovery and development of medical countermeasures for DoD’s Chemical and Biological Defense Program (CBDP). The workshop concluded with a small group discussion, limited to representatives from DTRA CB and its interagency funding partners, to identify investment opportunities. Key lessons learned from the workshop included the following:
- There are no standards for collecting, curating, preprocessing, or storing data for use in AI-and ML-based platform technologies.
- Experimental assumptions (e.g., operational definitions of data elements) can widely differ among research projects, thereby significantly affecting the predictive ability of ML algorithms.
- Current AI- and ML-related efforts across federal agencies could benefit from a more streamlined approach to investments.
DoD’s CBDP has disparate research projects funded by various stakeholders, which are necessary to gain diverse knowledge about known and emerging chemical and biological threat agents. The projects collect information on many data elements, but even if a handful of them were to share several data elements, definitions of the shared data elements would be often incongruous with each other. Currently, examining definitions for similarities requires human interpretation, which can be subjective. However, if the disparate research efforts were to collect a set of similar data elements sharing uniform definitions, then scientists could leverage AI and ML technologies to objectively and quickly compare data and identify trends. To enable this future characterized by expediency, DTRA CB will need to host additional workshops to explore data standardization so that in the future, CBDP-funded scientists will be able to access and manipulate data for ML.
Traditional drug-discovery processes can take years, yet new chemical and biological threat agents emerge more rapidly. A significant hastening of the drug-discovery timeline requires AI and ML; the two technologies can work faster than experimentation and human minds to learn characteristics of threat agents and identify their potential medical countermeasures. Successful employment of AI and ML technologies could make it possible to build a library of potential medical countermeasures for a broad collection of chemical and biological threat agents. The library would assure the health of the warfighter and defeat potential surprises — and might just disrupt the use of threat agents in warfare.
POCs: Sweta Batni, Ph.D., sweta.r.batni.civ@mail.mil; David Hone, Ph.D., david.m.hone2.civ@mail.mil
1. Stanford University. Professor John McCarthy: Father of AI; what is artificial intelligence? Stanford University website. http://jmc.stanford.edu/artificial-intelligence/what-is-ai/index.html. Accessed March 18, 2020.
2. Thompson W, Li H, Bolen A. Artificial intelligence, machine learning, deep learning and beyond: Understanding AI technologies and how they lead to smart applications. SAS website. https://www.sas.com/en_us/insights/articles/big-data/artificial-intelligence-machine-learning-deep-learning-and-beyond.html. Accessed March 18, 2020.
3. U.S. Department of Defense (DoD). Summary of the 2018 National Defense Strategy of the United States of America: Sharpening the American Military’s Competitive Edge. DoD website. https://dod.defense.gov/Portals/1/Documents/pubs/2018-National-Defense-Strategy-Summary.pdf. Accessed March 18, 2020.
4. Ibid.
| Date Taken: |
05.14.2020 |
| Date Posted: |
05.14.2020 11:39 |
| Story ID: |
369978 |
| Location: |
FORT BELVOIR, VIRGINIA, US |
| Web Views: |
425 |
| Downloads: |
1 |
PUBLIC DOMAIN
This work, Harnessing the Power of Machine Learning: A New Road to Rapid Drug Discovery, must comply with the restrictions shown on https://www.dvidshub.net/about/copyright.
Subscribe
