Back in March 2024, I wrote a blog about ARPA-H launching a new program called ADAPT (2024 blog here).
I noted then that Medicare patients were laboring under the outdated NCD 90.2, which assumes that major genomics on a cancer will only be done once - one and done.
From Genomeweb and elsewhere, here are some updates on progress with ADAPT, eighteen months later.
###
See a new subscription-based update at Precision Medicine Online by Alison Kanski - here.
###
See the ARPA-H home page for ADAPT, here.
See ARPA-H press release, May 2025, here.
See a listing of awardees with brief program descriptions, here.
See a press release from UNC Lineberger here. $28M.
See a press release from Inova here.
See a press release from Stanford here.
ARPA-H leadership spans two administrations - Acting Director Jason Roos PhD, who has held several HHS positions, came to ARPA-H in early 2024 and is still in the leadership role in August 2025. ARPA-H had a budget of $1.5B for FY2024 (here).
For the FY2026 budget proposal, several organizations may be merged (ASPR, ARPA-H, becoming ASHF), so it is more difficult to tease out an equivalent future budget for ARPA-H. However, I see a remark that ARPA-H will have a programmatic decrease of -$555M down to $945M total (p. 90). The new theme areas are proposed as, 'Chronic Disease," "America-Made Manufacturing," and "healthy Well being," "Healthcare efficiency and transparency," and "American Leadership."
###
AI CORNER
###
Using only the public links, I asked Chat GPT5 to draft an article for expert readers.
I have not proofed for details - so take it just as an example of AI writing style.
###
Here’s a structured ~1,500-word in-depth expert briefing on ARPA-H’s ADAPT program, written for a health policy and oncology expert audience.
Expert Briefing: ARPA-H’s ADAPT Program Targets Tumor Evolution with $142M Initiative
By [Your Name] – August 2025
Introduction
In May 2025, the Advanced Research Projects Agency for Health (ARPA-H) announced one of its most ambitious oncology initiatives to date: the ADvanced Analysis for Precision cancer Therapy (ADAPT) program. With up to $142 million in milestone-based funding, ADAPT aims to transform cancer care from static, biomarker-guided interventions into dynamic, continuously adaptive treatment frameworks that reflect the real-time biology of evolving tumors.
The program is emblematic of ARPA-H’s mandate to fund high-risk, high-reward research modeled after DARPA. Its ambition is not incremental improvement but rather a paradigm shift in precision oncology, particularly for patients with metastatic cancers where therapy resistance is inevitable and outcomes remain poor.
Why ADAPT Now?
Metastatic cancers remain one of the largest sources of mortality and cost in U.S. healthcare, accounting for more than 600,000 deaths annually. Nearly all metastatic tumors eventually develop resistance to systemic therapy. Traditional oncology practice rarely integrates real-time biological monitoring into decision-making: molecular profiling is typically performed once, prior to treatment, leaving clinicians blind to subsequent tumor adaptations.
The scientific case for ADAPT rests on several trends:
-
Tumor evolution is predictable: Recent advances in systems biology and mathematical modeling suggest resistant traits can often be anticipated.
-
Technology has matured: Next-generation sequencing, liquid biopsies, advanced imaging, and AI-driven multi-modal data fusion are now feasible at clinical scale.
-
Trial design is shifting: Adaptive clinical trial models are increasingly validated in other disease areas, demonstrating feasibility of “mid-course corrections.”
ARPA-H aims to integrate these strands into a single program that addresses the core challenge: how to continuously match therapy to a tumor’s moving biological target.
Program Architecture
ADAPT is organized into three technical areas (TAs), with 10 funded teams spanning academia, industry, and integrated cancer centers.
-
Therapy Recommendation Techniques (TA1)
-
Develop AI models, biomarkers, and computational frameworks that interpret longitudinal, multi-modal tumor data.
-
Goal: Identify resistant traits early, predict drug response, and recommend therapy switches.
-
-
Evolutionary Clinical Trial Design (TA2)
-
Test adaptive trial frameworks in breast, lung, and colon cancer.
-
Goal: Operationalize TA1’s predictions into real-world patient care.
-
-
Treatment and Analysis Platform (TA3)
-
Create a secure, cloud-based data ecosystem for real-time collaboration.
-
Goal: Provide shared infrastructure for clinicians, data scientists, and regulators.
-
Funding is milestone-contingent: continuation depends on aggressive progress toward predictive accuracy, trial enrollment, and clinical utility.
Technical Area 1: Therapy Recommendation Techniques
The five TA1 awardees reflect a cross-section of AI, computational biology, and biomarker development expertise.
-
Arizona State University – Tumor Ecology & Evolutionary Modeling
-
Focus: Map tumor resistance traits including gene activity, oncogenic state, immune evasion, and cellular evolvability.
-
Strategy: Borrow ecological models (e.g., predator-prey dynamics) to predict when resistance traits emerge and proliferate.
-
Significance: May provide a framework for “forecasting” tumor resistance before it is clinically evident.
-
-
Stanford University – Interpretable AI for Longitudinal Multimodal Data
-
Investigators: Olivier Gevaert, PhD, and Andrew Gentles, PhD.
-
Focus: Build machine learning models that integrate pathology, radiology, molecular, and clinical data.
-
Significance: Unlike “black box” AI, Stanford emphasizes interpretable models, ensuring clinical decision-makers understand why a therapy switch is recommended.
-
Quote: “This is the first adaptive trial where computational analysis is an intrinsic part of patient care,” said Gentles.
-
-
UC San Diego – Dynamic Biomarkers for Adaptive Therapy
-
Focus: Construct biomarkers that evolve in real-time alongside tumor biology.
-
Method: Continuous updating of predictive models using new biopsy and circulating tumor DNA (ctDNA) data.
-
Significance: Provides an infrastructure for serial biomarker re-interpretation, addressing the current bottleneck of static companion diagnostics.
-
-
MIT – Machine Learning for Therapy Optimization
-
Focus: Develop algorithms that synthesize genomic, imaging, and EHR data to recommend therapy sequences.
-
Significance: Bridges computational oncology with clinical informatics, potentially making ADAPT insights usable within existing health system data pipelines.
-
-
Brigham & Women’s Hospital – Foundational Models for Clinical-Genomic Data Fusion
-
Focus: Build multi-modal foundational models that integrate genomic, clinical, and treatment history data.
-
Role: Provide shared analytic tools to other TA1 teams, acting as a computational backbone.
-
Collectively, TA1 represents an attempt to bring machine learning into oncology practice at the same level of rigor as radiology or pathology.
Technical Area 2: Evolutionary Clinical Trials
Perhaps the most ambitious element of ADAPT is its commitment to evolutionary trial design: trials that adapt therapy while patients are on treatment, guided by TA1-generated biomarkers.
-
UNC Lineberger Comprehensive Cancer Center (Breast Cancer)
-
Award: Up to $28M.
-
Trial: EVOLVE (Evolutionary Clinical Trial for Novel Biomarker-Driven Therapies).
-
Enrollment: Up to 700 patients with metastatic breast cancer.
-
Methods: Serial biopsies, ctDNA monitoring, advanced imaging, and EMR integration.
-
Investigators: Lisa Carey, MD; Charles Perou, PhD; Ian Krop, MD, PhD; Eric Winer, MD; Antonio Wolff, MD.
-
Significance: Aims to generate predictive biomarkers where few exist today in breast cancer, particularly for resistance to endocrine and HER2-targeted therapies.
-
-
Beckman Research Institute of City of Hope (Lung Cancer)
-
Focus: Build a biomap of immunotherapy resistance mechanisms in non-small cell lung cancer.
-
Strategy: Validate new biomarker-driven therapies in near real-time, using immunogenomic profiling.
-
Significance: Tackles the growing clinical challenge of immune checkpoint inhibitor resistance.
-
-
UT MD Anderson Cancer Center (Colon Cancer)
-
Trial: Two-phase umbrella design for colon cancer.
-
Focus: Rapidly identify emergent resistant traits, integrate biomarker-guided drug assignments.
-
Collaborators: Inova Schar Cancer Institute among others.
-
Quote: “This is perhaps the most ambitious collaborative effort in precision oncology,” said Tim Cannon, MD, Inova.
-
Significance: Could create a new paradigm for adaptive colorectal trials, historically dominated by static trial designs.
-
Together, these trials test ADAPT’s central hypothesis: can continuously updated biomarker insights translate into longer survival and reduced toxicity in metastatic disease?
Technical Area 3: Treatment and Analysis Platform
Data infrastructure is the linchpin of ADAPT. Without a secure, interoperable, and scalable system, the adaptive framework would be impossible.
-
DNAnexus – Treatment and Analysis Platform (TAP)
-
Focus: Build a cloud-based data ecosystem linking clinicians, computational scientists, and trialists.
-
Features: Secure data sharing, regulatory compliance, real-time integration of TA1 biomarkers with TA2 trials.
-
Significance: Leverages DNAnexus’s commercial expertise in hosting genomics pipelines for scale.
-
-
Washington University in St. Louis – MAX System Integration
-
Focus: Extend TAP with Multi-modal Analysis with XNAT (MAX) tools.
-
Role: Provide advanced capabilities in managing imaging and clinical data, expanding TAP beyond genomics.
-
The TAP is also intended to generate public goods: anonymized aggregate datasets and algorithms will be made publicly accessible, creating an open science dimension to ADAPT.
Policy Implications
From a policy perspective, ADAPT raises several important issues:
-
Coverage & Reimbursement – Adaptive trials with mid-course treatment changes may challenge existing reimbursement frameworks (e.g., CMS coverage for companion diagnostics is tied to fixed, pre-treatment tests). ADAPT could pressure CMS and FDA to reconsider evidentiary standards.
-
Data Governance – With real-time cloud-based analysis, questions of HIPAA compliance, patient consent, and data security become central.
-
Clinical Guidelines – If ADAPT succeeds, NCCN guidelines may eventually need to integrate “adaptive biomarker monitoring” as a standard of care.
-
Regulatory Science – FDA’s role in regulating “evolving biomarkers” remains undefined; ADAPT could provide the first large-scale test case.
Outlook
Patient enrollment is expected within 12 months, with early results anticipated in 2–3 years. The program’s duration is set at six years, aligning with ARPA-H’s focus on achievable but transformative milestones.
The long-term goal is not simply better clinical trials, but a permanent change in how oncology is practiced:
-
from static biomarkers to dynamic, real-time monitoring,
-
from rigid protocols to adaptive trial frameworks,
-
and from siloed data to a shared national platform.
If successful, ADAPT could establish a blueprint for adaptive care models in other diseases, from infectious disease to neurodegeneration.
Conclusion
The ADAPT program represents ARPA-H’s most ambitious foray into oncology, combining the computational sophistication of Silicon Valley, the clinical trial expertise of NCI-designated centers, and the data infrastructure of leading biotech firms.
By aligning incentives through milestone-based funding, ARPA-H is forcing a level of cross-disciplinary integration rarely achieved in oncology. Whether ADAPT fulfills its promise will depend on whether predictive models can be translated into real-world trial outcomes – a challenge that has eluded precision oncology for decades.
For now, ADAPT stands as a test of ARPA-H’s core proposition: that disruptive, high-risk investments can accelerate U.S. biomedical innovation in ways incremental NIH funding cannot.
Word Count: ~1,540
?
