New around here? Start with the flagship paper

Dan Liu

Operator. Scientist. Bioengineer.

Flagship work

Lab on a Chip · Jul 2026 · Open access

A biomimetic microphysiological system predicts the impact of sepsis therapeutics on neutrophil-endothelial dynamics

Testing sepsis drugs in vessel-like channels with human cells should read out closer to clinical response than a static well plate.

About

I spent fourteen years on the industry side of drug development. The first half was close to the plant floor: process engineering, cGMP API manufacturing, production management. The second half was product and CMC, running global pharma product lines worth $12 to 15 million a year and leading teams of ten to twelve. I left industry as Associate Director for drug-device combination products at Lupin, serving as CMC lead on five complex assets, two of which advanced into clinical trials.

I kept hitting the same wall from the industry side: drugs that looked fine in animals and failed in people. So I went back for the science. I am now a Presidential Fellow at Temple University finishing a PhD in bioengineering, building the kind of models I wished I had as an operator: microphysiological systems that run human cells under flow, organoid and proteomics approaches, and network modeling to repurpose approved drugs. Sepsis is the proving ground. The tools are the point.

It is not a career change so much as a combination. The industry years are why the science is built the way it is: for reproducibility, throughput, and readouts a development team can actually use. Most platform scientists have never run a manufacturing line or filed a product. Most operators never train as bench scientists. I bring both to the same table.

Portrait of Dan Liu
Dan Liu gowned in a PAPR hood and containment coveralls in a pharmaceutical manufacturing gowning room
Gowned for containment — the plant-floor years.

The NAMs wave

Temple Bioengineering PhD candidate, graduating end of 2027 · Fourteen years in pharma CMC and combination products · Reading NAMs as drug-development tools: microphysiological systems, proteomics, AI drug repurposing.

The FDA is moving away from animal testing, and its roadmap leans into the disciplines I work in. NAMs, short for New Approach Methodologies, are human-relevant alternatives to animal testing: organ-on-chip systems, human cell-based assays, and in-silico models. The FDA Modernization Act 2.0 opened the door in 2022; in April 2025 the agency published a roadmap that starts with safety testing for monoclonal antibodies and aims to make animal studies “the exception rather than the norm” in preclinical safety testing within three to five years, naming those approaches among the paths forward.

March 2026 brought a draft validation framework from FDA’s drug center, CDER, for using NAMs in regulatory submissions; September 2026 brought a direct final rule rewriting “animal” to “nonclinical” across FDA drug regulations. I know the evidence bar from the submissions side: fourteen years in pharma taught me what regulators actually accept, and my bench work is built for that bar. The loop is the point: the chip generates human-relevant readouts, the proteomics and network layer ranks drugs from patient data, and each validates the other. I read the loop the way a CMC lead reads a process: what the readout is worth, what it costs, and whether a reviewer can trust it. Overlapping squarely with that shift is my own research:

Policy momentum, 2016–2026, backs the shift. See the full timeline below ↓

  1. Microphysiological systems

    A published chip that predicts therapeutic effect on vascular inflammation (Lab on a Chip, Jul 2026)

    A biomimetic microphysiological system that predicts how sepsis therapeutics affect neutrophil–endothelial dynamics: primary human endothelial cells and neutrophils, chemoattractant gradients, and physiological flow. Not a static well plate — something closer to a living vessel. Read the paper.

  2. In-silico and network modeling · AI drug repurposing

    Network models that prioritize repurposed drugs

    Neutrophil proteomics mapped into protein interaction networks, ranking FDA-approved drugs by how well each one normalizes a patient’s sepsis phenotype. Computation that is cheap, biology that is human — the in-silico layer of a NAMs-style evidence package. Frontiers in Immunology, Aug 2025.

  3. Platform transferability

    Sepsis is the proving ground, not the boundary

    The biology here travels: neutrophil–endothelial dynamics drive inflammation and vascular injury well beyond sepsis. The chip, the proteomics, and the network models form a reusable platform. Swap the disease context, keep the human-relevant machinery. That is what makes this a platform program rather than a single-disease project.

The policy arc

Ten years, one direction: from animal testing as the legal default to human-relevant methods as options under the law. The door is open. The evidence bar, fit-for-purpose validation, is the work ahead. This is the paper trail behind the shift my bench work is built for: FDA on the regulatory side, NIH on the funding side. Scroll sideways →

  1. Laying the rails · 2016–2020
  2. Milestone 1 of 14.

    Dec 2016

    The Cures Act builds the track

    The 21st Century Cures Act writes a formal Drug Development Tool qualification pathway into law — the legal rails that new kinds of tools, eventually organ-chips and AI models, would ride. Source

  3. Milestone 2 of 14.

    Dec 2017

    FDA bets on predictive toxicology

    The Predictive Toxicology Roadmap commits the agency to developing and qualifying new toxicology methods, naming organs-on-chips and mathematical modeling among them. Source

  4. Milestone 3 of 14.

    Nov 2020

    ISTAND opens the door

    FDA’s drug and biologics centers launch ISTAND, short for Innovative Science and Technology Approaches for New Drugs: a pilot opening Drug Development Tool qualification beyond biomarkers, to AI, digital health, wearables, and microphysiological systems. Source

  5. The door opens · 2022–2025
  6. Milestone 4 of 14.

    Sep 2022

    First non-animal tool enters the pathway

    FDA accepts the first ISTAND submission: a Membrane Proteome Array, a non-animal test for antibody off-target safety profiling. Acceptance shows the intake mechanism functions; qualification itself is the long road. Source

  7. Milestone 5 of 14.

    Dec 2022

    The animal testing mandate ends

    The FDA Modernization Act 2.0 is signed into law. Animal testing is no longer required by statute for new drugs; cell-based assays, microphysiological systems, and computer models are explicitly allowed as preclinical evidence. Source

  8. Milestone 6 of 14.

    Jan 2024

    NIH approves the alternatives program

    NIH’s Council of Councils approves Complement-ARIE, a Common Fund program to develop, standardize, validate, and qualify NAMs that complement — and in some cases replace — animal models. Source

  9. Milestone 7 of 14.

    Sep 2024

    The first organ-chip clears the first hurdle

    FDA accepts the Letter of Intent for Emulate’s Liver-Chip S1 for drug-induced liver injury risk assessment: the first organ-chip technology in the program. Acceptance into the program is a starting step, not an approval. Source

  10. Milestone 8 of 14.

    Apr 2025

    The phase-out roadmap

    FDA announces a plan to phase out animal testing requirements for monoclonal antibodies, with a stepwise path toward making animal studies the exception rather than the norm. Source

  11. Milestone 9 of 14.

    Apr 2025

    NIH prioritizes human-based research

    NIH announces an initiative to expand human-based methods and reduce animal use in funded research, and to establish the Office of Research Innovation, Validation, and Application (ORIVA) to coordinate validation. Source

  12. Milestone 10 of 14.

    Jul 2025

    NIH moves to end animal-only grant solicitations

    At a joint FDA–NIH workshop, NIH announces that all new funding opportunities moving forward should incorporate language on NAMs consideration, and that it will no longer seek proposals exclusively for animal models. Source

  13. Milestone 11 of 14.

    Jul 2025

    ISTAND becomes permanent

    FDA transitions ISTAND from pilot to a permanent Drug Development Tool qualification program. Qualifying novel tools is now a standing part of the framework. Source

  14. Milestone 12 of 14.

    Dec 2025

    Primate studies reduced; Modernization 3.0 advances

    FDA issues draft guidance, “Monoclonal Antibodies: Streamlined Nonclinical Safety Studies,” moving to reduce long-term nonhuman primate studies for certain monoclonal antibodies. The Senate passes the FDA Modernization Act 3.0 (S. 355) by unanimous consent to align the regulations with the 2022 law. Source

  15. The rulebook rewrites · 2026
  16. Milestone 13 of 14.

    Mar 2026

    The validation framework

    CDER issues draft guidance establishing a framework for validating NAMs for regulatory submissions — what evidence FDA needs to call a new method fit for purpose. Source

  17. Milestone 14 of 14.

    Sep 2026 Latest

    “Animal” leaves the rulebook

    A direct final rule replaces “animal” with “nonclinical” across FDA drug regulations. The rule does not mandate NAMs; it removes language that presumed animal studies. Which method fits still depends on qualification evidence. Takes effect February 4, 2027, barring significant adverse comments. Source

Key sources
  1. 21st Century Cures Act, Pub. L. 114-255, § 3011 (Dec 13, 2016) — Drug Development Tool qualification pathway.
  2. U.S. Food and Drug Administration, Predictive Toxicology Roadmap (Dec 2017). Roadmap (PDF)
  3. FDA CDER / CBER, Innovative Science and Technology Approaches for New Drugs (ISTAND) pilot program (launched Nov 2020; transitioned to a permanent qualification program, Jul 31, 2025). Program page · FDA Voices
  4. FDA Modernization Act 2.0, enacted Dec 29, 2022 as Sec. 3209 of the Consolidated Appropriations Act, 2023 (Pub. L. 117-328). Summary
  5. NIH Common Fund, Complement Animal Research In Experimentation (Complement-ARIE) program — concept approved by the NIH Council of Councils, Jan 25, 2024. NIH Director’s statement
  6. FDA, “Roadmap to Reducing Animal Testing in Preclinical Safety Studies” (Apr 10, 2025). Press release · Full roadmap (PDF)
  7. NIH, “NIH to prioritize human-based research technologies” (Apr 29, 2025) — initiative to expand human-based methods and establish ORIVA. News release
  8. FDA–NIH Workshop on Reducing Animal Testing (Jul 7, 2025) — NIH says new funding opportunities should incorporate NAMs consideration; will no longer seek proposals exclusively for animal models. Workshop page · NIH documentation (NIMH, Autumn 2025)
  9. FDA draft guidance, “Monoclonal Antibodies: Streamlined Nonclinical Safety Studies” (Dec 2025).
  10. FDA Modernization Act 3.0 (S. 355, 119th Congress) — passed by the U.S. Senate by unanimous consent, Dec 16, 2025. Congress.gov
  11. FDA CDER draft guidance, “General Considerations for the Use of New Approach Methodologies in Drug Development” (Mar 18, 2026). Draft guidance · Press release
  12. FDA direct final rule amending “animal” to “nonclinical” terminology in drug regulations, Docket No. FDA-2026-N-5347 (Sep 22, 2026; effective Feb 4, 2027). FDA summary
  13. Liver-Chip drug-induced liver injury study, Communications Medicine (2022) — 870 human Liver-Chips; the authors reported 87% sensitivity and 100% specificity for DILI in the study’s curated evaluation, prompting the 2024 ISTAND submission.

The direction is clear. And the operator side of me reads it a second way: throughput, standardization, and regulatory confidence are now the bottlenecks. I have stood on the submissions side of that divide for fourteen years; now I build from the other side.

Research

Concept figure for the organ-on-chip platform: chip fabrication and imaging, the digitized microvasculature design, in silico modeling, and genomics and proteomics analysis
The organ-on-chip platform, from concept to fabrication. Figure from Liu et al., Frontiers in Cellular and Infection Microbiology (2024).

Fourteen years in industry taught me where drugs fail: between the model and the patient. My lab work is about closing that gap. How do we test drugs on something closer to real human tissue? The work below runs from the biological case, to finding candidates, to testing them on the chip itself. Sepsis is where I proved the approach. The methods travel.

  1. Frontiers in Immunology

    Aug 2025 · Open access

    Analysis workflow: from sepsis-patient neutrophil proteomics to prioritized FDA-approved drug candidates

    Frontiers in Immunology · Aug 2025 · Open access

    Prioritizing FDA approved therapeutics for treating sepsis phenotypes: a network modeling approach based on neutrophil proteomics

    Why it matters. Repurposing approved drugs against matched phenotypes is faster and cheaper than new molecule programs.

    We mapped neutrophil proteomics from sepsis patients into protein interaction networks, then ranked FDA-approved drugs by how well each one could normalize a given neutrophil phenotype. The analysis surfaced three distinct targets: H2AC21 in the Hyperimmune phenotype, VTN in the Hybrid phenotype, and TRPV2 in the Hypoimmune phenotype.

    • AI drug discovery
    • proteomics
    • drug repurposing
    • sepsis

    Read the paper

  2. Frontiers in Cellular and Infection Microbiology

    Jan 2024 · Open access

    The organ-on-chip platform under the microscope, glowing with fluorescent tracer

    Frontiers in Cellular and Infection Microbiology · Jan 2024 · Open access

    The critical role of neutrophil-endothelial cell interactions in sepsis: new synergistic approaches employing organ-on-chip, omics, immune cell phenotyping and in silico modeling to identify new therapeutics

    Why it matters. The scientific case for the chip and the repurposing work that followed.

    A review of why the neutrophil-endothelium interaction is the right therapeutic target in sepsis, and why it takes organ-on-chip models, omics, and in silico modeling working together to find drugs. Sepsis accounts for more than 1 in 5 deaths worldwide, and animal models and traditional in vitro assays have repeatedly failed to translate.

    • sepsis
    • organ-on-chip

    Read the paper

Dan Liu beside a co-author in front of their Temple University poster on proteomic and in silico modeling of mouse endothelial cell dysfunction during inflammation Dan Liu talking through the poster with two visitors at the conference poster board
At a conference poster session: presenting the lab’s work on modeling endothelial cell dysfunction in inflammation.
Group photo of the Temple University biomicrofluidic lab in a conference room; Dan Liu is third from left
Dan Liu, third from left, with members of the Temple biomicrofluidic lab — the team behind the work.

Industry record

Fourteen years across manufacturing, product leadership, and CMC. Each role below shaped how I build science now: for translation, not just publication.

  1. Jul 2019 –
    May 2026

    Lupin Research Inc.

    Associate Director, Drug/Device Combination Products

    Joined as Senior Product Manager for development operations and rose to Associate Director for drug-device combination products. CMC lead on five complex assets, including inhalation and drug-device combinations. Advanced two complex generics into clinical trials. Managed ten-plus projects at once and supervised two junior PMs.

    Seven years from product manager to Associate Director: I learned what launch readiness demands, and that a platform has to survive device-grade scrutiny. I build like it will.

  2. Jan – Jul 2019

    Frontage Laboratories

    Director of Business Development, CMC

    Closed over $4M in CMC services in under seven months. Brought in 15+ new client accounts.

    Seven months inside a dozen sponsors’ CMC problems. You learn fast where development programs stall, and what a better early readout would be worth.

  3. Sep 2014 –
    Jan 2019

    Thermo Fisher Scientific (including Patheon)

    Global Pharma Product Manager

    Started as a Process Engineer in pharma API manufacturing at Patheon, before the Thermo Fisher acquisition. Owned global product lines worth $12–15M a year and led teams of ten to twelve.

    I started at the plant floor. That is where you learn what industry trusts: reproducibility, throughput, and a readout tied to a real decision.

  4. Sep 2012 –
    Sep 2014

    Cambrex

    Production Engineer / Project Manager

    Production engineering and project management in API manufacturing.

    A process is only as good as its controls. True for a reactor, true for a chip.

Notes

I am starting to write about the intersection I live in: running drug programs and building the science they depend on. Short pieces, no fixed schedule, first drafts in progress.

  1. The operator's view on NAMs

    What industry would actually need to adopt organ-on-chip in place of animal models: throughput, standards, and trust.

    • organ-on-chip
    • NAMs
  2. AI drug discovery's validation problem, from someone who builds the assays

    Computational predictions are cheap. The bottleneck is a believable bench model to test them in.

    • AI drug discovery
    • organ-on-chip
  3. Phenotype first, molecule second

    The repurposing lesson from sepsis: match the patient subgroup before you pick the drug.

    • drug repurposing
    • proteomics

The running diary version of this lives at Field Notes — one paper at a time, with the industry translation. Until these essays land, shorter takes go on LinkedIn.

Contact

Talk to me about chips, CMC, or sepsis.

The fastest way to reach me is LinkedIn. I am always up for a conversation about microphysiological systems, drug repurposing, combination products, or what it takes to move a molecule from the lab to the patient.