ARPA-H Project Awardees
The ARPA-H Mission Office Innovative Solutions Openings (ISOs) and Open Broad Agency Announcement (BAA) provide funding for research that aims to improve health outcomes across a wide range of patient populations, communities, diseases, and conditions. These projects focus on transformative ideas for health research breakthroughs or technological advancements.
Awards made from the ISOs and Open BAA are generally in the form of contractual agreements. Exact award amounts are dependent upon meeting milestones typical of the ARPA-H process. As of March 2024, ARPA-H is no longer accepting submissions for the Open BAA solicitation, but ARPA-H will continue to review and consider solution summaries and proposals submitted under the Open BAA before it closed. Currently, ARPA-H will primarily use Program-Specific and Mission Office ISOs to advertise and accept submissions for its programs and projects.
ARPA-H is pleased to announce the following awardees:
V-CARES: Human-centered Design of an Ethical Evaluation Strategy for Chatbot Hallucinations in Health care
The goal of the project led by Vanderbilt University Medical Center is to create the Vanderbilt Chatbot Accuracy and Reliability Evaluation System (V-CARES) to effectively and efficiently detect hallucinations, omissions, and misaligned values from large language model (LLM) responses in the healthcare domain. The project aims to improve the quality of LLM-based Chatbot systems so they can serve as trustworthy, transparent, and accurate sources of information and guidance on health-related topics for the general public, patients, and their caregivers. In addition to reducing hallucinations and omissions, the project aims to create a generalizable evaluation process and technology to improve the quality of the output responses and ensure they are consistent with users’ values and expectations. Immediate use cases include two distinct and prevalent mental health disorders: Major depression and generalized anxiety disorder.
ALICE: Assessing LLM Integrity for Clinical Engagement
The goal of the project led by the University of Southern California is to develop an automated and efficient approach to medical chatbot evaluation with evaluation components designed to automatically generate questions that reflect a patient-community’s needs and scoring components that reflect common challenges (e.g., hallucinations, omissions, miscalibrated linguistic certainty) as well as pragmatic needs specific to the community. To support this development, USC will engage with stakeholders to understand their information needs and how they are addressed. Immediate use cases include pediatric infectious disease and cystic fibrosis.
MiRIT: Micro-Radiolabeling for Imaging and Therapy
This project aims to develop personalized radiopharmaceuticals for PET imaging-based diagnosis and targeted treatment for multiple cancers. This will be achieved through the computer-aided design of custom ligands targeted to cancer cell membrane proteins and through the development of two table-top devices. The first device will be capable of producing very small doses of radiopharmaceuticals on-site at the patient clinic through a single push of a button. The second device will permit parallelized labeling under varying reaction conditions and using varying ligands/radioisotopes, enabling fast-track development of diagnostic and treatment radiolabeled compounds. The resulting targeted imaging and treatment will enable personalized diagnostic PET imaging and radiotherapy, the latter of which could easily be extended to resource-limited clinics and hospitals via a low-dose micro-radiolabeling system developed through the project.
PAIL: PhotoAcoustic Imaging technology for diagnostic Lung assessment
There is a critical need to develop minimally invasive in vivo imaging technologies capable of providing microscopic assessment and identification of malignancy at early stages of lung cancer, as early detection and treatment is essential to optimize patient survival. The Northeastern team will develop a photoacoustic imaging (PAI) system, with a miniature (1.5 mm diameter), flexible, and disposable probe that is fully compatible with the requirements of bronchoscopy. The PAI system is expected to achieve an unprecedented axial and lateral resolution of ~27 μm and ~100 μm at cm-scale depths. The proposed system has the potential to become the first ever commercial low-cost endobronchial PAI (EB-PAI) platform and will enable 3D microscopic visualization of lung nodules, reducing the risk, burden, and cost associated with biopsy. Though lung nodule assessment will be the proof-of-concept application for this project, the technology itself can be easily adapted to impact many other areas of clinical practice (e.g., cardiology, nerve detection).
NEBULA: NExt-generation Biomanufacturing ULtra-scalable Approach
NEBULA aims to create a ‘cells made in a box’ small-footprint, low human resource required cell expansion platform for GMP manufacturing of autologous cell-based therapies. The project aims to unlock personalized, affordable regenerative medicine treatments for the diverse U.S. population via an autonomous biomanufacturing system scaling the production of personalized induced pluripotent stem cells (iPSCs). If successful, the project can lay the foundation for broader innovation and access to treatment for diseases like Parkinson’s, heart failure, spinal cord injury, and age-related macular degeneration, among other conditions, affecting nearly 30 million Americans. NEBULA aims to facilitate the research and development of personalized, affordable treatments, significantly reducing the national health burden of these conditions.
DAGCAP: Democratized, AI-Guided Chart Abstraction Platform
Chart curation is a major bottleneck for oncology clinical practice and research. Existing automated approaches are insufficient to address this bottleneck as important patient data are stored as natural language rather than structured data. Current tools for automated processing of natural language remain inaccurate, require substantial human oversight, and lack oncology-specific knowledge. DAGCAP is a modular, interoperable, comprehensive end-to-end workflow aimed at decreasing chart curation effort by 90% while maintaining human-level accuracy. Users will be able to precisely define data variables using a guided, interactive process, and these variables will be mapped to commonly used data models as well as those defined or imported by users. The proposed research spans 1) Tool Features - building out the software components that will enable drastic improvement in chart curation capabilities; 2) Data Types & Semantics - introduce and train DAGCAP to understand and map between a variety of data types, data models, and enable data analysis and use of downstream tools; and 3) Use Cases - representative of real-world problems that we will use to develop, test, and improve DAGCAP.
IndiPHARM: Individual Metabolome and Exposome Assessment for Pharmaceutical Optimization
The project will deliver a platform to optimize drug regimens for patients, especially those that have metabolic conditions or disorders. The platform will be designed to be used at the point of care to assist in decision making regarding the efficacy and precision medicine principles of existing drugs. It will consist of two main toolkits: predictive modeling software and a fully referenced high resolution mass-spectrometry (HRMS) workflow. The predictive software will leverage patient data from blood samples to model their xenobiotic metabolism, considering factors such as genomic background. The HRMS workflow, with its quick turnaround time and cost-effectiveness, will provide accurate measurements of patients' blood samples for the metabolism of therapeutics and pharmacokinetics/pharmacodynamics (PK/PD) modeling. The workflow and modeling toolkits will also have the capability to conduct parallel and integrated analyses at population scale, to help clinicians and product developers decipher the wide range of metabolism that affects the delivery of effective therapeutics. This comprehensive approach will enhance our understanding of how drugs act in different patient populations, as well as how they affect individuals on a more personalized level, in line with precision medicine approach.
Safe and Explainable AI-enabled Decision Making for Personalized Treatment
The project is focused on the design and implementation of AI-based clinical decision support systems for personalized treatment and management recommendations. The proposed research spans three areas. First, it addresses AI foundations—problems in trustworthy medical AI, such as integrating medical domain knowledge in learning models effectively, making recommendations of AI algorithms explainable to clinicians, and establishing worst-case safety guarantees. The second focus is on AI systems—infrastructure to facilitate development of explainable models suitable for integration into clinician workflows. Thirdly, the project looks at AI use cases—representative clinical challenges that span inpatient and outpatient use cases, including prediction of in-hospital cardiac arrest, timely diagnosis and prediction of the need for intervention for sepsis, and prediction of response to neoadjuvant or adjuvant chemotherapy for breast cancer patients. To execute this agenda, the team brings together clinicians and researchers with expertise spanning AI, biostatistics, data science, and machine learning.
SUPPLI: Strategic Utilization of Pharmaceutical Product Location and Identification
The SUPPLI project addresses supply shortages for the key chemicals required to manufacture pharmaceuticals by creating a tool guided by artificial intelligence (AI) that (1) calculates the risk of supply side shortages and prioritizes efforts based on demand, (2) reverse engineers the relevant molecules to derive the starting ingredients, and (3) predicts appropriate manufacturing facilities, conditions, and protocols for manufacturing the chemical components in US based facilities. There is no current published framework for quantifying the interdependence of the manufacturing process from regulatory starting materials (RSMs) to final product. SUPPLI aims to de-risk the supply chain with a focus on the domestication of RSM manufacturing with a targeted goal of creating a product that will optimize pharmaceutical lifecycles, improve process traceability, and provide sustainable process designs that provide more efficient access to pharmaceuticals with fewer supply chain disruptions. It will also aid in the optimization of costs of manufacturing, which may decrease the downstream cost burden on patients and allow more accessible therapeutics to a great number of Americans.
MASCOT: Manufacturing Agile and SCalable Organoid Tumor models
One roadblock when developing cancer treatments is the lack of preclinical models that faithfully recreate the tumor microenvironment (TME). Innovative 3D tumor models can replicate the TME, leading to better diagnosis of the disease conditions, surgical planning or drug selection for the patient. The UIUC team will use advanced manufacturing strategies—Industry 4.0—to develop a revolutionary platform that will consistently and reproducibly produce three-dimensional (3D) tumor models of any cancer type, at scale. The platform will exert closed-loop control over the production of tumor organoids while generating a digital thread that documents the status of the tumor models at every stage. The effort includes significant automation for processing, development of high-speed and chemically detailed non-destructive imaging to characterize the models, and the creation of an AI-based Model Predictive Control to dictate growth conditions at each stage of the tumor model development. The prototype manufacturing technology will produce consistent and verified tumor models at a rate of ≥10 per day. This agile manufacturing technology will provide a turnkey solution for validated 3D models for any solid tumor, including cancers from individual patients, cancers prevalent in underserved and minority communities, or rare cancers unaddressed by current technologies.