The Next Map of the Human Body

Multiomics is where genomics, AI, and the next wave of precision medicine start to collide.

The first human genome was supposed to explain everything.

That was the pitch. Decode the three billion letters. Find the bad genes. Fix disease at the source. A clean story. A beautiful story. Also incomplete.

Because DNA is not destiny. DNA is closer to a script locked in a drawer. What matters is which pages the cell reads, which lines it ignores, which proteins it actually builds, which metabolites pile up like exhaust, which immune cells show up late, which microbes whisper from the gut, and where all of this happens inside real tissue.

That is multiomics.

Not one layer of biology. Many layers, stacked together until the body stops looking like a parts list and starts looking like a city at night: roads, electricity, water pressure, traffic jams, police reports, restaurant receipts, and heat maps all blinking at once.

My view: multiomics today feels a lot like genomics did 10 to 15 years ago.

The science is real. The commercial plumbing is uneven. The early revenue lines do not yet capture the size of the eventual market. Some public stocks will turn out to be monsters. Some will be disasters with beautiful slides. A few boring suppliers may make more money than the glamorous pure plays. That is usually how platform shifts work.

And this one matters because medicine is moving from static genetic clues toward living biological context. AI is becoming good enough to model that context. Quantum computing is not the engine today, but over a 5 to 20 year horizon it could become a second acceleration layer for chemistry, simulation, and optimization.

What multiomics really means

Genomics asks: what is written in the DNA?

Transcriptomics asks: which genes are being read into RNA right now?

Proteomics asks: which proteins are present, modified, folded, degraded, and doing work?

Metabolomics asks: what small molecules and chemical byproducts reveal the cell’s current state?

Epigenomics asks: which DNA switches are open or closed without changing the underlying genetic code?

Microbiome analysis asks: what organisms are living in and around us, and how are they changing immune, metabolic, or inflammatory behavior?

Single cell analysis asks the question bulk biology used to hide: which exact cells are doing what?

Spatial biology adds the missing map: where are those cells inside the tissue, who are their neighbors, and what conversations are happening at the tumor border, in the inflamed gut, or inside the aging brain?

Each layer is useful. Together, they are different in kind. Disease rarely respects the boundaries of a textbook. Cancer involves DNA, but the mutation is only the opening clue. Alzheimer’s involves protein aggregates, but the surrounding cellular neighborhood matters too. Autoimmune disease involves immune cells, but the trigger and tissue context decide the damage. Metabolic disease involves glucose, but the real body is a feedback loop.

For decades, medicine worked like a detective arriving after the crime. A patient is sick. A symptom appears. A blood test or scan catches a clue. Doctors infer what may have happened upstream.

Multiomics tries to move the camera earlier. It wants to see the molecular weather before the storm breaks.

That is why I think the category is still under-imagined.

The genomics movie already played once

Go back to 2010 or 2012 and genomics had the same weird feeling.

The Human Genome Project was old news by then. Sequencing costs were collapsing. Illumina was becoming the default machine inside the lab. Cancer sequencing was starting to move from research into clinical decision making. Consumer DNA had a lot of cultural heat. Every serious hospital wanted to understand precision medicine, even if most of them had no idea how to operationalize it.

The skeptical case sounded reasonable: sequencing was expensive, reimbursement was messy, interpretation was hard, privacy was unresolved, and doctors did not need another 200 page PDF full of variants of unknown significance.

All true.

It still became one of the most important life sciences platform shifts of the last 15 years.

Illumina is the cleanest historical example. Around 2010, the stock traded near $30 on a split adjusted basis. By 2015, it was around $180. By 2021, near the peak of the genomics and life sciences tool boom, it briefly traded above $500. Then the story broke. Growth slowed, competition rose, the Grail deal became a governance and regulatory mess, and the stock fell hard. The lesson is not “buy every sequencing stock.” The lesson is sharper: the platform was real, the adoption curve was real, and the equity path still punished anyone who ignored valuation, execution, and cycle risk.

Thermo Fisher and Danaher tell a quieter story. They were not sexy genomics pure plays. They were diversified life sciences infrastructure companies. Yet they benefited from the same underlying wave: more sequencing, more sample prep, more bioprocessing, more lab automation, more analytical workflows. From the early 2010s to the 2021 life sciences peak, both became much larger and more strategically important. The boring toolchain often compounds while the market argues about the headline technology.

Exact Sciences is another useful case. It turned molecular screening into a commercial product with Cologuard, then pushed deeper into cancer testing. The stock had enormous runs and painful drawdowns. Guardant Health came public later and gave investors a more direct liquid biopsy story. It also showed the other side of the curve: the science can be impressive while profitability, reimbursement, competition, and trial evidence remain brutal.

Then came the more specialized names. PacBio had moments of real excitement around long-read sequencing and then long stretches of investor pain. 10x Genomics came public in 2019 as one of the best known single cell and spatial biology names, rocketed during the 2020-2021 life sciences enthusiasm, and then collapsed with the rest of the high-multiple tool universe. Again: the technology did not become irrelevant. The stock did not get a free pass.

That is the right mental model for multiomics.

A platform can be inevitable and still destroy capital in the wrong vehicle. A platform can be early and still create huge winners in the supply chain. A platform can look like a research tool for years before it becomes a clinical or pharmaceutical workflow. By the time revenue is obvious, the best entry prices may already be gone. By the time the hype is loudest, the risk may already be hiding in plain sight.

This is why I do not want to treat multiomics as a cute subcategory of genomics. Genomics gave medicine the street map. Multiomics is trying to add traffic, weather, power consumption, emergency calls, surveillance footage, and the weird smell coming from the alley.

That is a bigger data problem. It is also a bigger business problem.

The field is early in the way that matters

If you listen only to conference-stage language, you might think multiomics has already transformed healthcare. It has not.

But “not transformed healthcare yet” is not the same as “not investable” or “not important.” That distinction matters.

In research, multiomics is already part of serious biology. Single cell RNA sequencing changed how scientists look at tissue. Spatial transcriptomics and spatial proteomics let researchers see biology without grinding everything into molecular soup. Large biobanks, tumor atlases, clinical sequencing programs, imaging archives, and electronic health records are turning biology into a layered dataset.

In oncology, rare disease, immunology, cardiovascular research, aging, metabolic disease, and drug discovery, the old one-marker-at-a-time model feels increasingly primitive.

The clinic is slower.

Genomic tests are commercial. Some expression based tests are commercial. Some cancer assays already combine DNA, RNA, methylation, protein markers, pathology, imaging, or clinical data into one decision. But full-stack multiomics for routine diagnosis is still held back by cost, sample handling, standardization, reimbursement, regulation, and the boring reality that hospitals are not built like computational biology labs.

That gap is not a reason to dismiss the space. It is where the companies will be made.

The science wants thousands or millions of measurements. The clinic wants one answer that is accurate, reimbursed, explainable, and delivered before the physician loses patience.

Someone has to turn a mountain of molecular noise into a clinically useful sentence.

This patient needs drug A, not drug B.

This tumor is likely to recur.

This target is not worth five years of R&D.

This subgroup of patients, invisible in the old trial design, is where the drug actually works.

That is where the money should concentrate. Not in data for data’s sake. In decisions.

AI is the translation layer

Multiomics without AI is like owning a library where every book is written in a different language and half the pages are missing.

A human researcher can understand a pathway. A team can analyze one dataset. But multiomics quickly becomes too wide, too sparse, too noisy, and too high dimensional for old methods alone. The comparison is no longer gene A versus disease B. It is millions of weak signals across cells, tissues, time, drugs, clinical records, imaging, and outcomes.

AI is the translation layer.

Machine learning can integrate messy data types, impute missing layers, cluster patient subgroups, identify biomarkers, predict drug response, and prioritize targets. Deep learning can learn representations of cells the way language models learn representations of words. Foundation models for biology are the natural extension of this idea.

One clear example is scGPT, published in Nature Methods in 2024. The researchers trained a generative model across more than 33 million cells and showed it could support cell type annotation, batch integration, multiomic integration, perturbation response prediction, and gene network inference. That does not make scGPT a finished commercial product. It shows the direction: biology is moving from handcrafted analysis toward pretrained biological representations.

In plain English: AI is learning a grammar of cells.

If that grammar gets reliable, the consequences are large.

Drug discovery changes because teams can ask better questions earlier. Biomarker discovery changes because models can see patient subtypes hidden inside noisy averages. Clinical trials change because inclusion criteria can be built around biology instead of blunt demographic categories. Diagnostics change because molecular context can follow the patient instead of sitting in a one-time lab report.

The important 5 to 20 year question is not whether every model works this year. Many will fail. Wet labs are merciless. Correlations die. Batch effects bite. Hospital data is ugly. But that is exactly what happened in genomics: interpretation lagged measurement, then workflows caught up, then commercial products emerged.

The investment question is simple: who owns scarce biological data, repeatable workflows, customer access, and a path from prediction to validation?

That is a much smaller list than “everyone with AI in the deck.”

Quantum computing is the far-horizon accelerator

Over the next few years, quantum computing probably will not be the main driver of multiomics. The heavy lifting will come from sequencing, single cell tools, spatial biology, mass spectrometry, proteomics platforms, cloud infrastructure, GPUs, lab automation, and AI models trained on better biological data.

But over 5, 10, or 20 years, quantum belongs on the map.

Molecules are quantum systems. Drug discovery and biology eventually run into problems where classical approximations are expensive, incomplete, or painfully slow: electronic structure, binding energy, reaction pathways, conformational states, protein-ligand interactions, materials for instruments, and certain optimization problems in experimental design. If fault tolerant quantum computing matures, it could become a serious tool for chemistry and simulation.

That does not mean quantum replaces multiomics platforms. It means quantum may deepen the model of what the multiomics platforms are measuring.

Imagine a future workflow. Multiomics identifies a disease state at high resolution. AI proposes the pathway, the target, the subgroup, and the likely response. Quantum accelerated chemistry helps simulate the molecular interaction or reaction space with more fidelity than classical tools can manage alone. Robotic labs test the best candidates. The results loop back into the model.

That is not tomorrow morning. It is not a revenue forecast for IonQ, Rigetti, D-Wave, IBM, Microsoft, Alphabet, or NVIDIA. It is a direction of travel.

And direction of travel matters in platform markets.

The more practical near-term link is hybrid computing. NVIDIA’s CUDA-Q work with Google Quantum AI is a useful signal because it shows the strange loop forming: classical accelerated computing helps design and simulate quantum processors, and those quantum systems may eventually help simulate chemistry. IBM, Microsoft, Alphabet, IonQ, Rigetti, and D-Wave all sit somewhere on that map. Investors should still separate a quantum platform company from a multiomics revenue company. But I would not dismiss the connection just because the first commercial impact is indirect.

The next biology stack may be layered like this: assays create the data, AI interprets the data, accelerated computing trains the models, and quantum eventually attacks the hardest simulation edges.

That is a 20 year sentence. I think it is worth writing down now.

The public watchlist

I would group the public companies into four baskets.

The first basket is the toolmakers: Illumina (ILMN), 10x Genomics (TXG), Thermo Fisher Scientific (TMO), Danaher (DHR), Agilent (A), Waters (WAT), QIAGEN (QGEN), PacBio (PACB), and Standard BioTools (LAB).

These companies sell the picks and shovels. Illumina is still central to sequencing despite its bruised stock chart. 10x Genomics is one of the names most closely tied to single cell and spatial biology, with Chromium, Visium, and Xenium. Thermo Fisher and Danaher are broader, less pure, and deeply embedded in labs. Agilent and Waters matter because mass spectrometry, chromatography, biomolecular analysis, and analytical workflows are part of the proteomics and metabolomics stack. QIAGEN sits in sample prep, molecular assays, NGS workflows, and bioinformatics. PacBio gives long-read exposure. Standard BioTools, especially through the SomaLogic combination, gives exposure to proteomics and high-parameter biology, though execution risk is real.

The second basket is diagnostics and clinical data: Tempus AI (TEM), SOPHiA GENETICS (SOPH), Exact Sciences (EXAS), Guardant Health (GH), and Labcorp (LH).

This basket is closer to the patient. Tempus is one of the more direct public expressions of the thesis: multimodal clinical, molecular, imaging, and text data with AI layered on top. Its 2024 filing described more than 7.3 million de-identified patient records, over 900 million documents, more than 1.3 million records with matched clinical and genomic information, and more than 260,000 with full transcriptomic profiles. Those are company-reported figures, not magic. But they explain why the market pays attention.

SOPHiA GENETICS is another direct name, with a cloud-native platform built around data-driven medicine and multimodal analysis across genomics, radiomics, clinical, biological, and digital pathology data. Exact Sciences and Guardant Health show how molecular signals can become reimbursed clinical products, and also how expensive the road can be. Labcorp is broader and less pure, but scale, samples, ordering relationships, and clinical lab infrastructure matter.

The third basket is computational biology and AI drug discovery: Schrödinger (SDGR), Recursion (RXRX), and AbCellera (ABCL).

Schrödinger is more physics based computational chemistry than multiomics, but its combination of molecular simulation and machine learning belongs on the adjacent map. Recursion is closer to the high-dimensional biology thesis, with phenomics, transcriptomics, proteomics, patient-centric data, and machine learning inside its operating system. AbCellera is focused on antibody discovery and development rather than broad multiomics, but it fits the same pattern: industrialized biological data plus computation plus a platform business model.

The fourth basket is infrastructure: NVIDIA (NVDA), Alphabet (GOOGL), Microsoft (MSFT), IBM (IBM), IonQ (IONQ), Rigetti (RGTI), and D-Wave Quantum (QBTS).

NVIDIA is the most immediate infrastructure beneficiary because modern biological AI runs on accelerated computing. Alphabet and Microsoft matter through cloud, AI research, health data tooling, and quantum research, though neither is a pure biology stock. IBM has long-running quantum and enterprise research exposure. IonQ, Rigetti, and D-Wave are quantum watchlist names. Their connection to multiomics is indirect today. Over a 10 to 20 year horizon, that indirect exposure may become more interesting if quantum chemistry and hybrid AI workflows mature.

The trick is not to buy a theme. The trick is to understand where the theme touches revenue, margins, data advantage, and customer behavior.

A ticker is not a thesis. But sometimes a theme tells you where to look before the thesis is obvious.

The frontier may leak out before it shows up in public revenue

Some of the most important signals are not clean public-stock stories yet.

In the measurement layer, a few names are public or tucked inside larger companies: Oxford Nanopore trades in London, PacBio trades on Nasdaq, Olink now sits inside Thermo Fisher, and SomaLogic sits inside Standard BioTools. They matter because they show where the instrument race is going: longer reads, cheaper reads, better proteins, better spatial context, and less friction between sample and insight.

The private-company watchlist is different: Element Biosciences, Ultima Genomics, Parse Biosciences, Vizgen, Scale Biosciences, Isomorphic Labs, Generate:Biomedicines, Insilico Medicine, and Owkin.

These companies are pushing either the measurement layer or the AI-biology layer before the story becomes obvious in public filings.

Private leaders matter because they reveal the frontier early. They also create temptation. Fewer filings. Fewer audited numbers. More room for heroic narratives. The right posture is not cynicism. It is aggressive curiosity with a calculator in your hand.

The risks are real, but they are not the whole story

Multiomics can generate gorgeous plots that do not improve outcomes. That is the first risk.

Biology is physical. Samples degrade. Tissues vary. Labs use different protocols. A model trained on one dataset may stumble on another. The boring operational details are not boring; they are the moat.

Reimbursement can also break the dream. A diagnostic can be scientifically elegant and commercially miserable if nobody pays for it.

Then there is stock selection. Large diversified companies may benefit from the trend but barely move because multiomics is only one small part of revenue. Small pure plays may offer more upside to the theme but carry financing risk, adoption risk, and violent drawdowns.

The 2010s genomics lesson is useful here. The wave was real. Sequencing did become central. Precision oncology did grow. Clinical genomics did spread. Tool companies made money. Diagnostics companies built products. Infrastructure companies compounded.

And still, plenty of shareholders got hurt.

That is not a reason to avoid the map. It is a reason to read the map correctly.

My read

Multiomics is not a single product category. It is a direction of travel.

Medicine is moving from one-dimensional tests toward layered biological context. AI is moving from generic pattern recognition toward domain-specific biological models. Drug discovery is moving from artisanal trial and error toward industrialized data loops. Quantum computing sits farther out, but if chemistry and simulation become quantum-assisted over the next 10 to 20 years, the biology stack gets another gear.

The most interesting companies will connect four things: hard-to-get biological data, reliable measurement workflows, models that improve with scale, and a commercial path into decisions people pay for.

That is why I keep coming back to the city-at-night image.

For a long time, genomics gave us the street map. Valuable, yes. Revolutionary, yes. But a street map does not tell you where traffic is frozen, where the power is out, where the hospital is overwhelmed, or where a fire is about to start.

Multiomics is the attempt to turn on the lights.

And once the lights are on, the real competition begins.