This article is a run-through of organoids and organs-on-chips in drug discovery!
Drug discovery has a translation problem. We keep pretending that the next mouse study, the next 2D assay, the next “good enough” toxicity package will somehow predict the messy reality of human biology. Sometimes it does. Often it does not. And when it does not, everyone pays: patients, investors, scientists, founders, pharma teams, and the poor project leader who has to explain why the beautiful animal data collapsed in Phase II.
This is why organoids, organs-on-chips, tissue chips, microphysiological systems, assembloids, spheroids, and related human model systems matter. Not because they are cute miniature organs in a dish. Not because they make nice conference slides. They matter because they challenge a deeply expensive habit: outsourcing human predictivity to models that were never human in the first place.
The promise of this field is simple: build more human-relevant biology earlier, make better decisions earlier, and stop pushing translational failure downstream.
That sounds obvious, but as we all know, it’s also extremely hard.
The Promise: Human Biology Before Human Trials
Organoids are three-dimensional, self-organising cellular systems derived from stem cells or tissue samples. Their strength is biological richness. They can preserve aspects of cellular diversity, architecture, patient genetics, and disease phenotype that flat cultures often either erase or never had to begin with.
Organs-on-chips are engineered microfluidic systems where living cells are cultured under controlled flow, mechanical forces, and tissue interfaces. Their strength is environmental control. They can add perfusion, stretch, barrier function, shear stress, concentration gradients, and sometimes multi-organ crosstalk.
Together, they represent a shift from cells in a petri dish or flask to something more biologically relevant: human tissue under conditions that begin to resemble physiological conditions.
The FDA now explicitly promotes lab-grown human organoids and organ-on-chip systems as part of its move toward reducing animal testing, particularly where human-relevant models can reveal toxicities that animals may miss. FDA’s New Alternative Methods Program also aims to support methods that replace, reduce, and refine animal testing while improving predictivity in nonclinical testing. NIH NCATS describes tissue chips, also called organs-on-chips, as human-cell-based systems designed to mimic organs such as heart, kidney, and lung for faster and more effective drug discovery and -effect testing.
This is an operating-model shift that speeds up processes, provides better go/no-go decisions and erodes the need for laboratory animals.
A better model produces better biology. It improves portfolio management. It changes go/no-go discipline. It changes how early a research team can see toxicity, poor exposure, failed barrier penetration, immune interaction, organ crosstalk, and patient-segment effects.
In other words: the model becomes a decision tool. That is where the real value is.
A Short History: From Clever Biology to Industrial Translation
The organoid story did not start as a CRO service or a platform pitch. It came from developmental biology, stem cell biology, and the realisation that cells can self-organise when given the right signals and environment. Modern intestinal organoid work accelerated after Sato, Clevers, and colleagues showed that intestinal stem cells could form long-term, self-renewing crypt-villus-like structures in 3D culture. That changed the field because it showed that complex tissue-like organisation could be maintained outside the body.
The organ-on-chip story took a more engineering-driven path. The Wyss Institute describes Donald Ingber’s organ-chip work as emerging from microengineering and the concept of recreating organ-level functions in living, microfluidic devices. (Wyss Institute) The famous “living, breathing lung-on-a-chip” work showed that a small device containing human lung and blood-vessel cells could mimic key aspects of lung function and response. (Harvard Gazette)
So we have two traditions now converging:
- Organoids: biology first, complexity from self-organisation
- Chips: engineering first, physiology from controlled microenvironments
The obvious next step is not to argue which is better. That is just theatre. The next step is to combine them where the question requires it. Organoids and organs-on-chips in drug discovery work together.
A cancer organoid may be enough for a drug-response screen. A kidney chip may be better for transport and toxicity. A BBB chip may be essential for CNS delivery. A brain organoid-on-chip may be needed when disease biology, neuronal activity, maturation, perfusion, and exposure all matter.
The tool must follow the question. Not the other way around.
Current Strengths: Where These Models Already Earn Their Keep
The strongest case today is not that organoids and chips replace everything. They most definitely do not. The strongest case is that they reduce the incidence of bad decisions.
1. Better human relevance in early discovery
Traditional 2D cultures are useful, cheap, scalable, and often necessary. But they are also brutally simplified. Organoids can preserve patient-specific disease features and tissue architecture, while chips can recreate flow, barrier function, and mechanical forces.
This matters in oncology, liver toxicity, kidney toxicity, gut biology, inflammatory disease, rare disease, and increasingly CNS.
2. Earlier toxicity signals
Drug-induced liver injury, kidney toxicity, cardiotoxicity, seizure liability, and BBB-related failures can destroy programs late. Late is expensive. Late is painful. Late is often avoidable.
Organ-on-chip systems are especially relevant where toxicity depends on exposure, metabolism, transport, or multicellular interaction. FDA’s acceptance of an organ-on-chip technology into its ISTAND Pilot Program for drug-induced liver injury shows that these systems are moving from impressive science toward regulatory qualification pathways. (U.S. Food and Drug Administration)
3. Better disease modelling for patient subgroups
Patient-derived organoids and iPSC-derived systems make it possible to ask questions that animal models often handle poorly: Which genetic background matters? Which tumour subtype responds? Which toxicity is patient-specific? Which rare disease biology cannot be modelled properly in rodents?
This is stratification at its best.
4. More useful failures
A failed experiment in a human-relevant model can be valuable if it explains why a candidate should not advance. A fast NO-GO decision is of grand commercial benefit. The best model is not the one that gives beautiful data. The best model is the one that prevents futile spending.
Current Weaknesses: The Hard Parts Nobody Should Hide
This field has developed a bad habit of overselling itself. That is always dangerous as a day of reckoning comes to all projects. Organoids and organs-on-chips are not yet plug-and-play replacements for animal studies, clinical trials, or human biology. They are decision aids. Some are excellent. Some are fragile. Some are expensive toys with nice fluorescence images.
The main weaknesses are clear.
1. Reproducibility and Industrialisation
Brain organoids currently face critical hindrances, as identified in Nature Communications, specifically heterogeneity, size variation, cellular stress, and poor reproducibility. However, recent research in Nature indicates that industrialising high-throughput generation can reduce this heterogeneity, enabling robust drug screening. Overcoming these defects requires a shift toward manufacturing discipline, incorporating automation, batch-to-batch control, and standardised quality gates, to ensure these models function as reliable decision-grade systems rather than unrepeatable lab crafts.
2. Maturity
Many organoids resemble fetal or developmental tissue more than adult tissue. That can be perfect for neurodevelopmental disease. It can be problematic for adult neurodegeneration, chronic toxicity, metabolism, and ageing biology.
3. Missing systems
Organoids often lack vasculature, immune compartments, innervation, endocrine interaction, stromal architecture, and full mechanical context. Chips can add some of this, but complexity increases cost, reduces throughput and risks increase the noise in your data. You can build a more realistic system. You can also build a system nobody can run twice. Organoids and organs-on-chips in drug discovery still have a way to go.
4. Throughput and standardisation
Pharma does not adopt platforms because they are elegant. Pharma adopts platforms when they are reliable, documented, transferable, and decision-relevant. That means SOPs, controls, reference compounds, batch records, assay windows, acceptance criteria, inter-lab transfer, and boring documentation. As always: boring wins.
5. Regulatory acceptance
The FDA’s alternative-methods work is about qualification, guidelines, applied research, and filling information gaps. (U.S. Food and Drug Administration) The industry still needs defined contexts of use. A model validated for liver DILI screening is not automatically valid for chronic neuroinflammation, BBB transport, or seizure liability.
Context of use will remain the gatekeeper.
Why CNS May Become the Most Important Battleground
CNS drug development is where hope goes to be tested brutally. The biology is hard. Human brain development differs from that of rodents. The blood-brain barrier blocks many therapeutic concepts. Neurodegenerative diseases unfold over years. Patient heterogeneity is enormous. Clinical endpoints can be slow, noisy, and expensive. A candidate can look compelling for far too long before reality arrives.
This is exactly where better human models can create value. Killing weak hypotheses earlier and strengthening the ones that deserve clinical risk is what the world needs.
Just keep this in mind: Nothing replaces the patient.
The CNS opportunity has four attractive zones.
First: blood-brain barrier and blood-CSF barrier models. These are not academic ornaments. They are central to delivery. A 2026 review describes BBB and blood-CSF barrier organ-on-chip systems as important tools because these interfaces not only protect the CNS but also pose major drug-delivery and disease-modelling challenges. Microfluidic flow, 3D architectures, and human stem-cell-derived cells help recapitulate the in vivo environment more closely. (Springer)
Second: brain organoids for disease biology. Brain organoids can model neurodevelopmental disease, patient-specific genetics, glioma invasion, and some neurodegenerative mechanisms. The best future use is not a generic “brain in a dish”. It is disease-specific, endpoint-specific and carefully benchmarked for use.
Third: electrophysiology and functional readouts. CNS models become more useful when morphology is connected to function. Neuronal firing, network activity, synaptic behaviour, calcium dynamics, MEA readouts, and seizure liability matter because CNS failure is often functional before it is visibly structural.
Fourth: integrated neurovascular and immune systems. The future CNS model will not be a lonely brain organoid floating in a dish. It will likely combine region-specific neural tissue, BBB or neurovascular units, microglia or immune components, controlled perfusion, and functional monitoring. Nature Reviews Bioengineering makes the practical point: brain-on-chip complexity must be tailored to the research question. (Nature)
That last point is critical. More complexity is not automatically better.
A stupidly complex model is still stupid.
The winning CNS platforms will be those that match complexity to decision value.
Ten Companies to Watch
This is not a ranking of any kind. It is a practical operator’s map of companies and platforms worth watching across organoids, organs-on-chips, CNS models, liver and kidney safety, multi-organ systems, and industrialised 3D biology mixing organoids and organs-on-chips in drug discovery.
1. Emulate
Emulate is one of the most visible organ-chip companies, with liver, kidney, intestine, lung, and other chip systems. Its Liver-Chip S1 was accepted into FDA’s ISTAND Pilot Program, an important regulatory-facing signal for the field. (Emulate)
Why it matters: regulatory credibility and pharma familiarity.
2. MIMETAS
MIMETAS builds OrganoPlate-based systems and offers perfused 3D models, including kidney models with proximal tubule, distal tubule, and glomerulus elements. Its approach is attractive because it pushes organ-on-chip biology toward higher-throughput plate workflows. (mimetas.com)
Why it matters: scale and workflow compatibility.
3. CN Bio
CN Bio’s PhysioMimix platform supports organ-on-chip work for human-specific efficacy, ADME, and safety data. The company launched PhysioMimix Core in 2025 as an all-in-one system intended to support single-organ, multi-organ, and higher-throughput configurations. (CN Bio)
Why it matters: multi-organ use and translational pharmacology.
4. Hesperos
Hesperos develops Human-on-a-Chip systems and operates as a biotechnology CRO focused on safety and efficacy testing. Its positioning is especially relevant for teams that need service-based access rather than building chip capability internally. (Hesperos Inc.)
Why it matters: service model for translational decisions.
5. TissUse
TissUse’s HUMIMIC platform connects organ models through microfluidic channels and on-chip pumping, enabling multi-organ interaction studies for drugs, chemicals, cosmetics, and food additives. (tissuse.com)
Why it matters: systemic interaction and multi-organ crosstalk.
6. InSphero
InSphero is strong in 3D microtissues, especially liver safety. Its 3D InSight Human Liver Microtissues are assay-ready spheroids designed for population-relevant liver response studies, and a published FDA NCTR collaboration benchmarked liver-toxicity prediction across 152 approved drugs. (InSphero)
Why it matters: standardised 3D liver biology and safety screening.
7. Molecular Devices and Cellesce
Molecular Devices acquired Cellesce to expand access to patient-derived organoid technology and large-scale organoid manufacturing for applications including drug screening. (moleculardevices.com)
Why it matters: organoid scale-up, imaging, automation, and assay infrastructure.
8. Crown Bioscience and HUB Organoids
Crown Bioscience offers patient-derived organoid services based on HUB Organoids technology, with emphasis on oncology drug development and standardised 3D assays. (crownbio.com)
Why it matters: oncology PDO translational screening.
9. 28bio, formerly AxoSim
28bio, linked to AxoSim’s neurotechnology lineage, focuses on human brain organoids for neurotoxicity testing. Its CNS-3D Brain Organoids are being evaluated to predict clinical seizure liability using small molecules with documented human clinical outcomes. (28bio.com)
Why it matters: CNS functional risk, especially seizure liability.
10. NETRI
NETRI develops neurons-on-chip and neuro-organ-on-chip systems. Its collaboration with Axion BioSystems connected NETRI’s neuro-organ-on-chip devices with MEA systems for applications including Alzheimer’s disease, Parkinson’s disease, ALS, peripheral neuropathic pain, gut-brain axis, and neurotoxicity. (axionbiosystems.com)
Why it matters: neurofunctional readouts and CNS disease modelling.
Where the Future Strengths Lie
The next level will not come from prettier organoids. It will come from decision-grade systems.
Five future strengths are especially important.
1. Context-of-use validation
Every serious platform needs a clear claim.
Not “we model the brain.”
Say: “We predict seizure liability for this class of small molecules under these assay conditions with this acceptance criterion.”
Not “we model the liver.”
Say: “We identify DILI risk for compounds with this metabolic profile in this concentration range.”
Vague platforms impress investors for a short time. Specific platforms win procurement and regulatory trust.
2. Hybrid biology and engineering
Organoids bring cellular complexity. Chips bring environmental control. AI brings pattern detection. Imaging brings morphology. MEA brings function. Multi-omics brings the mechanism.
The future is not one model. It is a qualified stack.
3. Better CNS maturation and vascularisation
For CNS, the key bottlenecks are maturity, reproducibility, vascularisation, immune integration, and functional relevance. Better neurovascular-unit models, BBB chips, microglia-containing organoids, vascularized scaffolds, and long-term cultures will matter.
But again: do not worship complexity. Use the minimum complexity required to answer the decision question.
4. Industrial reproducibility
The winners will have manufacturing discipline. Batch-to-batch control. Cryopreservation. Reference standards. Automation. Image analysis pipelines. Defined quality gates.
In brain organoids, recent work shows that high-quantity generation can reduce heterogeneity and enable drug screening, which is exactly the kind of industrialisation the field needs. (Nature)
5. Commercial models that fit pharma behaviour
A platform can be scientifically superior and commercially awkward. That is still a problem.
Pharma needs internal transfer options, CRO service options, clear pricing, training, validation packages, reference data, and procurement simplicity. A technology that requires heroic PhD wizardry for every run will remain a specialist tool.
No heroics. Systems.
What Needs to Happen to the Industry
A pragmatic checklist.
- Define contexts of use and stop selling “human biology in a dish.” Sell validated decisions.
- Build reference-compound libraries. Every platform needs positive controls, negative controls, borderline cases, and known clinical failures.
- Standardise quality gates. Size, viability, marker expression, functional activity, barrier integrity, morphology, transcriptomics where relevant, and acceptance criteria must be documented.
- Improve transferability. A model that works only in the inventor’s lab is not a product. It is a craft.
- Connect biology to commercial decisions. The output must affect candidate selection, dose strategy, safety margin, patient segmentation, delivery design, or clinical trial design. Otherwise, it is just an expensive decoration.
- Use CNS models with discipline. BBB models for delivery. Brain organoids for disease biology. MEA and functional systems for neuronal activity. Integrated chips for neurovascular and multi-compartment questions. Match the model to the decision.
- Build regulator-facing evidence early. The FDA direction is encouraging, but qualification depends on evidence, not enthusiasm. (U.S. Food and Drug Administration)
- Stop pretending animals will disappear tomorrow. The next era is not “animal models are dead.” The next era is model portfolios. Human systems, animal systems, computational systems, clinical data, and RWE are used where they are strongest.
My Takeaways
Organoids and organs-on-chips are not the end of preclinical risk. They are the beginning of a more honest conversation about it.
Their promise is not that they make drug discovery easy. Nothing does.
Their promise is that they can make drug discovery less lazy, less animal-dependent, less surprised by human biology, and less willing to push failure downstream.
CNS may become the proving ground. The field desperately needs better ways to model human brain biology, BBB transport, seizure liability, neurodegeneration, neuroinflammation, and patient-specific mechanisms before expensive clinical bets are made.
But the winners will not be the companies with the nicest organoid pictures. The winners will be the ones that turn human-relevant biology into validated, reproducible, decision-grade systems.
That is what needs to happen in order to take this to the next level.
Not more hype.
Better gates. Better models. Better decisions.
At Biosector, we are always open to discuss your projects. We are based in Tokyo and help overseas companies conduct business with Japan’s pharma industry. Warmly welcome to contact us for a meeting info@biosector.jp
Comments are closed