Kingy AI
Subscribe, 2.1M subscribersWork with Kingy
AI News

Cascade: How a 100-Million-Cell AI Project Could Transform Drug Discovery

October 31, 2026 AI News
AI-powered drug discovery

A New Chapter in AI-Powered Medicine

What if scientists could predict how the human immune system responds to a drug before spending years testing it?

That possibility is moving closer to reality.

On October 29, 2026, Parse Biosciences and Nomic Bio announced Cascade, an ambitious biological data project designed to help artificial intelligence understand the human immune system.

The scale is remarkable.

Cascade aims to combine 100 million individual cell RNA profiles with more than 11,000 protein profiles. Together, these measurements could give researchers a much clearer picture of how immune cells react to different substances.

And this isn’t some small laboratory experiment.

Pharmaceutical heavyweights GSK, Pfizer, and Sanofi have joined the initiative as foundational partners. Biotechnology companies Graph Therapeutics and Turbine are also participating.

The objective is straightforward but extraordinarily difficult: create better biological data so AI systems can predict immune responses more accurately.

For pharmaceutical researchers, that could mean identifying promising drug targets earlier, understanding complicated immune reactions, and making smarter decisions about which treatments deserve further investigation.

However, Cascade isn’t a newly approved treatment or a finished AI system that can independently discover medicines.

It’s something more fundamental.

It’s an attempt to build the biological foundation that future AI-powered drug discovery systems will need.

And that distinction makes the project particularly interesting.

Why Drug Discovery Still Takes So Long

Developing a medicine involves far more than finding a chemical that produces an interesting reaction.

Scientists must understand how that substance interacts with the human body.

Will it affect the intended biological target? Could it trigger an unwanted immune response? Will its effects remain consistent across different patients?

Answering those questions takes time.

Researchers typically move through laboratory experiments, preclinical studies, and multiple phases of human clinical trials before a medicine can receive regulatory approval.

Many candidates fail somewhere along that journey.

AI promises to make parts of this process more efficient.

Machine learning systems can analyze enormous biological datasets, identify patterns, and help researchers prioritize promising experiments.

But there’s a problem.

Even sophisticated AI models struggle when the information used to train them doesn’t capture enough of the underlying biology.

Imagine teaching someone to predict tomorrow’s weather using temperature readings alone.

Those readings would help. Yet without information about wind, humidity, and atmospheric pressure, important pieces would remain missing.

Something similar happens when AI models study immune cells.

They may understand which genes become active without fully capturing the signals those cells send to their neighbors.

Cascade wants to address that missing information.

Meet the Companies Behind Cascade

Two biotechnology specialists are combining their expertise to make the project possible.

Parse Biosciences focuses on single-cell sequencing, a technology that allows researchers to examine gene activity in individual cells.

The company, which is part of QIAGEN, operates GigaLab, a high-throughput facility designed to process enormous numbers of cells.

Nomic Bio specializes in proteomics.

While genomics and transcriptomics help researchers understand genetic information and gene activity, proteomics examines proteins, the molecules responsible for countless biological functions.

Nomic’s technology can measure large numbers of proteins released by cells.

That capability matters because immune cells communicate extensively through molecular signals.

By combining their technologies, the companies hope to create a dataset that captures both internal cellular activity and external protein communication.

Think of it as studying a busy airport.

Parse can help researchers understand what’s happening inside individual terminals.

Nomic can reveal the messages traveling between them.

Together, those perspectives provide a more complete view of how the entire system operates.

The partnership also brings specialized laboratory infrastructure into a field increasingly dominated by AI ambitions.

Because impressive algorithms still need reliable experimental evidence.

The 100-Million-Cell Ambition

Cascade’s headline figure deserves attention.

One hundred million cells.

That’s the planned scale of its single-cell RNA sequencing dataset.

Each profile will help describe which genes are active within an individual cell under particular experimental conditions.

Researchers can then compare those profiles to investigate how different immune cell populations respond to drugs and signaling molecules.

But Cascade isn’t simply collecting cells and placing their measurements into a giant database.

The project is designed around controlled biological experiments.

Scientists will expose immune cells to different substances and observe how their responses change.

The announced experimental framework includes more than 150,000 perturbation measurements.

A perturbation is an intentional change introduced into a biological system.

For example, researchers might expose immune cells to a particular signaling molecule or drug compound.

They can then examine what happens.

Which genes become more active? Which become less active? What proteins do the cells release?

By repeating these experiments across different conditions, scientists can assemble a detailed map of immune behavior.

The project’s planned scale could make that map especially useful for training and evaluating AI models.

However, the 100-million-cell figure represents the dataset’s intended scope.

It should not be confused with a claim that every planned measurement was already completed when Cascade was announced.

Why Proteins Are the Missing Piece

Here’s where Cascade becomes particularly interesting.

Many existing large-scale biological datasets focus heavily on RNA.

RNA measurements can reveal which genes a cell is expressing.

That information provides valuable clues about what the cell is doing.

But gene expression doesn’t tell the entire story.

Cells also produce and release proteins.

Some of these proteins act as messengers, influencing the behavior of nearby cells and coordinating immune responses.

Scientists call the collection of proteins released by cells the secretome.

Understanding the secretome is especially important when studying inflammation, immune regulation, and responses to treatments.

Consider two immune cells exposed to the same drug.

Their RNA profiles might reveal similar changes in gene activity.

Yet the amounts and combinations of proteins they release could differ.

Those differences may influence how other cells respond.

Cascade aims to connect these two kinds of information.

Instead of examining gene expression and protein signaling separately, researchers will be able to study their relationships under matched experimental conditions.

That could provide AI systems with a more informative picture of immune biology.

And better information is often the first requirement for better predictions.

Watching the Immune System Change Over Time

Biology doesn’t stand still.

A cell might react to a substance within minutes, produce different signals hours later, and eventually return toward its original state.

A single measurement cannot capture that entire process.

Cascade addresses this challenge through time-resolved experiments.

Researchers will collect information at multiple points after exposing immune cells to different substances.

This approach should help reveal how biological responses develop.

Some signals may appear quickly.

Others might emerge later.

Certain reactions could be temporary, while others persist.

Those differences matter when researchers investigate medicines designed to influence immune activity.

A drug that produces a useful initial response might trigger less desirable effects afterward.

Conversely, a treatment that appears ineffective during an early measurement could produce meaningful changes later.

Cascade’s design aims to capture these changing patterns.

Its original announcement described experiments using immune cells from 12 healthy donors and examining at least 14 immune cell types.

That provides a controlled foundation for studying immune behavior.

Still, healthy-donor experiments cannot fully reproduce the biological complexity of patients living with different diseases.

Further studies will be necessary to establish how well models trained on these measurements perform in disease-specific settings.

How AI Could Use Cascade’s Data

Imagine an AI system trained to recognize patterns in millions of immune cell responses.

Researchers introduce information about a drug candidate.

The model predicts which cellular pathways might change and what protein signals could follow.

Scientists could then test those predictions in the laboratory.

That’s one potential application of Cascade.

The dataset is intended to support predictive models of immune biology, sometimes described as virtual cell models.

These systems attempt to simulate aspects of cellular behavior using computational methods.

They don’t create perfect digital copies of living cells.

Instead, they learn relationships from experimental measurements and use those relationships to estimate responses under different conditions.

Cascade could help researchers develop and evaluate such systems.

For example, scientists might train a model using some experimental conditions and test whether it can predict results from conditions it hasn’t encountered.

They could also examine whether the model performs consistently across different donors.

That process is essential.

An AI system that merely memorizes its training data offers limited scientific value.

Researchers need models capable of making useful predictions about new biological situations.

Cascade’s combination of RNA measurements, protein data, and multiple timepoints creates opportunities to investigate those capabilities.

The ultimate objective is not to eliminate experiments.

It’s to make computational predictions more useful in deciding which experiments to conduct.

Why Pfizer, GSK, and Sanofi Are Interested

AI-powered drug discovery

The participation of major pharmaceutical companies gives Cascade additional significance.

GSK, Pfizer, and Sanofi have extensive experience developing medicines and investigating complex biological processes.

All three operate in areas where understanding immune responses can be important.

Their involvement signals commercial interest in better datasets for computational biology.

But why would pharmaceutical companies support a project focused on collecting biological measurements?

Because drug discovery depends heavily on understanding how treatments affect living systems.

A promising compound might influence several biological pathways.

Some effects could be beneficial.

Others might complicate development.

More informative experimental data could help researchers investigate those relationships earlier.

For pharmaceutical organizations already investing in computational research, datasets like Cascade may provide valuable training and evaluation material.

Importantly, the companies’ participation doesn’t prove that Cascade will produce successful medicines.

Nor does it establish that participating organizations have already achieved superior clinical outcomes using the dataset.

The announcement identifies them as foundational partners.

The practical benefits will depend on the quality of the completed data and how effectively researchers use it.

Still, their involvement suggests that the industry sees value in building stronger biological foundations for AI.

Could Cascade Help Researchers Study Cancer?

Cancer treatment provides an interesting example of where better immune modeling might eventually prove useful.

The immune system plays a central role in identifying and responding to abnormal cells.

Some cancer treatments attempt to strengthen or redirect immune activity.

However, immune responses can vary considerably between patients.

A treatment that benefits one person may produce a different result in another.

Researchers need to understand the complicated relationships among immune cells, signaling molecules, and therapeutic interventions.

Cascade could contribute foundational knowledge about some of those relationships.

Its measurements may help scientists investigate how different immune cell populations respond to selected substances.

That information could eventually support research into cancer immunotherapy.

But an important limitation remains.

Cascade’s announced foundational experiments use immune cells from healthy donors.

They aren’t a comprehensive representation of tumors, cancer patients, or the complex environments surrounding malignant cells.

Consequently, researchers would need additional disease-specific measurements before drawing conclusions about particular cancer treatments.

Cascade may help build the tools.

It does not, by itself, demonstrate that those tools can predict cancer treatment success.

What About Autoimmune Diseases?

Autoimmune conditions present another possible research application.

These diseases involve immune responses that mistakenly target the body’s own tissues.

Examples include rheumatoid arthritis, lupus, and multiple sclerosis.

Scientists studying these conditions often investigate how immune cells communicate and how inflammatory signals influence disease activity.

Cascade’s protein measurements could provide useful information about some of those signaling mechanisms.

Its time-resolved design may also help researchers investigate how responses change after exposure to immune-modulating substances.

That could support the identification of biological pathways worth exploring as potential drug targets.

However, the project’s initial healthy-donor dataset is not a substitute for studying patients with autoimmune diseases.

Disease-specific samples and experiments would be necessary to establish whether predictions remain useful in those settings.

The distinction matters because immune biology is extraordinarily context-dependent.

A response observed in a controlled laboratory experiment may behave differently inside a patient.

Cascade’s potential lies in providing a stronger starting point for research.

Turning that foundation into clinical benefits will require considerably more work.

The Technology Making Cascade Possible

Projects of this scale depend on specialized laboratory infrastructure.

Parse Biosciences plans to use its GigaLab facility for single-cell RNA sequencing.

The facility is designed for high-throughput analysis, allowing researchers to process very large numbers of cells.

Meanwhile, Nomic Bio will use its Omni 1000 platform to measure secreted proteins.

The platform can examine more than 1,000 proteins across experimental samples.

Combining those technologies is central to Cascade’s design.

The challenge isn’t simply collecting a large amount of information.

Researchers must ensure that measurements from different technologies can be meaningfully connected.

They also need consistent experimental procedures, reliable sample tracking, and rigorous quality controls.

Otherwise, AI models could learn misleading patterns.

For example, differences caused by laboratory procedures might appear to be biological differences.

That would undermine the dataset’s usefulness.

The project therefore represents an effort in experimental design as much as an investment in data collection.

Its success will depend on how accurately researchers can connect cellular changes with protein signaling across different conditions and timepoints.

Why Bigger Data Doesn’t Automatically Mean Better AI

One hundred million cells sounds impressive.

But size alone cannot guarantee scientific progress.

An enormous dataset containing inconsistent or biased measurements might produce unreliable predictions.

Researchers must consider which biological conditions the data represents.

They also need to understand what the dataset leaves out.

Cascade’s initial framework uses samples from 12 healthy donors.

That provides opportunities to study variation among donors, but it cannot capture the full diversity of human immune responses.

Age, genetics, disease status, medication exposure, and many other factors can influence immune behavior.

The laboratory environment also differs from the conditions inside a living human body.

These limitations don’t make Cascade unimportant.

They establish the boundaries within which its results should be interpreted.

Strong AI research requires careful evaluation.

Scientists will need to determine whether models trained on Cascade can predict unseen experimental outcomes accurately.

They must also investigate whether those predictions remain useful when researchers introduce new biological contexts.

The most meaningful measure of success won’t be the number of cells in the database.

It will be whether the information helps scientists make better, independently verifiable predictions.

An Important Question: Who Gets Access?

Cascade also raises questions about access to valuable scientific data.

The companies describe the project as a licensable dataset.

That means it is being developed for access through commercial data partnerships rather than announced as an unrestricted public resource.

This business model could help support the substantial costs associated with generating and maintaining high-quality biological measurements.

It could also provide participating pharmaceutical organizations with access to specialized information.

However, licensing arrangements may affect which researchers can use the data.

Academic laboratories, smaller biotechnology companies, and independent research groups may face different access conditions.

Those details matter because scientific progress often benefits from broad participation and independent evaluation.

A dataset becomes more valuable when researchers can test models, reproduce findings, and identify limitations.

Cascade’s commercial structure therefore deserves attention alongside its scientific ambitions.

The announcement establishes that licensing will be available.

It doesn’t establish that the entire dataset will be freely accessible.

What Happens Next?

Cascade’s announcement marks the beginning of an ambitious effort, not the completion of one.

The companies reported that data generation was underway.

The next challenge is producing the planned measurements while maintaining consistent experimental quality.

Researchers will then need to determine how effectively the resulting information supports predictive AI models.

Several questions remain open.

Can models trained on the dataset predict protein responses under previously unseen conditions?

Will the measurements help researchers identify promising drug targets?

Can the experimental framework expand successfully into disease-specific research?

And perhaps most importantly, will better predictions eventually translate into measurable improvements in drug development?

Those answers will require evidence.

For now, Cascade provides a substantial new direction for studying immune biology with AI.

Its combination of cellular measurements, protein signaling, and time-resolved experiments addresses an important limitation in many existing biological datasets.

The project also demonstrates how biotechnology companies and pharmaceutical organizations are working together to improve the information available to AI systems.

That’s a meaningful development, even before any new medicines emerge from the research.

The Bigger Picture: Teaching AI the Language of Biology

Artificial intelligence has already demonstrated impressive capabilities in language, images, programming, and scientific analysis.

But living systems present a different challenge.

Biology involves countless interacting processes operating across different scales and times.

Understanding those processes requires more than recognizing patterns in isolated measurements.

Researchers need information that connects different layers of biological activity.

Cascade represents an effort to create exactly that kind of foundation.

By combining gene-expression measurements with protein signaling, the project aims to help AI systems develop a richer understanding of immune responses.

The pharmaceutical industry is paying attention.

Major drugmakers have joined the initiative, and the participating biotechnology companies are building infrastructure to support its ambitious scale.

Yet the most exciting possibilities remain future goals rather than established achievements.

Cascade hasn’t demonstrated that AI can reliably predict every immune response.

It hasn’t eliminated the need for laboratory experiments or clinical trials.

And it hasn’t produced a newly approved medicine.

What it could provide is something that scientific progress depends on: better evidence.

Because before AI can help researchers predict how complex biological systems will behave, it needs reliable examples of those systems in action.

Cascade is attempting to provide millions of them.

And if that effort succeeds, the benefits could extend well beyond a single dataset.

It could help researchers build AI models that ask better questions, guide more informative experiments, and eventually contribute to more effective approaches to discovering medicines.

That’s a future worth investigating.

Sources