AI, Biomarkers and Drug Design - Mexico Business News

STORY INLINE POST

In the previous article, we explored how AI x multi-omics help us move from molecular chaos to predictive cancer care.

But if AI can help us understand cancer, what happens when we use it not only to interpret the disease, but to design the next generation of treatments against it?

For decades, drug development has looked like searching for a hidden door inside a dark labyrinth. Scientists tested thousands of molecules, followed biological clues, failed repeatedly, and sometimes — after years, billions of dollars, and hundreds of candidates — found one drug strong enough to reach patients.

Artificial intelligence does not remove the labyrinth, but it gives us a map. And in cancer, where each tumor can behave like a different biological universe, that map may become one of the most powerful tools medicine has ever had.

Welcome to the next frontier of precision oncology: AI-assisted biomarker discovery and drug design.

The real promise of AI in oncology is not simply faster drug discovery. It is reducing biological uncertainty — helping us understand which target matters, which patient is most likely to respond, and which therapy deserves to move forward.

Cancer Is Not Only a Disease — It Is a Pattern

Cancer leaves clues everywhere: In DNA mutations, RNA expression, abnormal proteins, altered metabolism, immune evasion and resistance mechanisms.

The problem is not that cancer gives us too little information. The problem is that it gives us too much.

A tumor molecular profile (TMP) can reveal hundreds of genomic alterations. Transcriptomics can show thousands of genes turned on or off. Proteomics can reveal which pathways are actually active. Clinical records show what happened with therapies that were actually used.

Separately, these are fragments. Together, they become a biological fingerprint and AI helps us read it.

From Biomarkers to Biological Vulnerabilities

A biomarker is not just a laboratory result, it’s a clue.

Sometimes it tells us that a tumor may respond to a drug, like EGFR mutations in lung cancer.

Sometimes it warns us that a treatment may not work.

And sometimes it reveals a hidden vulnerability, a pathway the tumor depends on, and sometimes even abuses, to survive (aka oncogenic addiction)

This is where AI becomes valuable.

Machine learning can analyze thousands or millions of data points and identify patterns that humans would miss. It can ask:

- Which mutation actually drives the tumor?

- Which RNA expression pattern predicts resistance?

- Which patients look biologically similar even if their cancers started in different organs?

In other words, AI helps turn molecular noise into therapeutic direction.

The Lock, the Key — and Now the Blueprint

In one of our previous articles, we described how monoclonal antibodies work as a lock-and-key model.

If a cancer cell has a specific lock (alteration), we can design a key (monoclonal antibody) to target it.

But AI takes us one step earlier.

Instead of only asking, “Which key fits this lock?” AI asks: Which lock matters most — and can we design a better key?

We are no longer only matching existing drugs to known targets. We are beginning to discover new targets, design new molecules, and predict which patients may benefit before a clinical trial even begins.

That is the logic behind smarter, biomarker-driven clinical trials: fewer patients, better filters, and a higher probability of meaningful response.

How AI Helps Design New Cancer Drugs

AI can support cancer drug development across several layers.

First, it can identify new targets by comparing tumor multi-omic data with healthy tissue, looking for molecular dependencies that cancer cells need but normal cells can survive without.

Second, it can help design new molecules. Instead of manually testing compound after compound, generative AI can propose structures predicted to bind a specific target.

Third, it can predict how proteins fold and interact. This matters because proteins are not flat letters on a page — they are three-dimensional machines. Their shape determines their function and advances like AlphaFold have made protein structure prediction dramatically faster and more accessible.

Fourth, AI can help select patients for clinical trials by identifying biomarker-enriched populations — patients whose tumors are more likely to respond because of their molecular profile.

This is crucial.

A drug does not fail only because the molecule is bad. Sometimes it fails because it was tested in the wrong patients.

From Trial-and-Error to Trial-with-Reason

Traditional oncology drug development often starts broad.

A therapy is tested in a tumor type: lung, breast, colon, pancreas. Precision oncology teaches us that tissue origin is only part of the story.

Two lung cancers can be molecularly different diseases.

Breast cancer and gastric cancer may share HER2 amplification.

Colorectal cancer and melanoma may both reveal immune-related biomarkers.

AI can help reorganize oncology around biology rather than anatomy alone.

Imagine developing a new drug for a pathway activated across multiple tumors.

Such was the case with pembrolizumab, one of oncology’s clearest examples of biology-driven treatment across tumor types. Pembrolizumab does not target the tumor by organ; it blocks PD-1, an immune checkpoint receptor on T cells, helping restore immune recognition of cancer cells that use PD-L1/PD-L2 signaling as a cloaking mechanism.

Its tumor-agnostic approvals in MSI-H/dMMR and TMB-high cancers showed that sometimes the relevant question is not where the tumor started, but which biological vulnerability it carries.

Also instead of enrolling patients only by organ, AI-supported biomarker discovery could help identify the subgroup most likely to respond — regardless of where the tumor started.

That means smaller, smarter, faster trials.

Not easier trials. Smarter ones.

Real-World Signals: This Is Already Happening

This is not theoretical.

Companies like Insilico Medicine, Recursion, Exscientia, and others are building AI-first or AI-supported drug discovery platforms.

Insilico reported that its generative AI-discovered and designed candidate entered Phase II clinical trials, a milestone for AI-driven drug development.

Big Pharma is also moving aggressively and does not see AI as a toy. Takeda recently entered an AI drug-discovery partnership with Insilico Medicine with potential value up to $600 million, while Insilico has also announced major AI-driven drug discovery agreements with companies such as Eli Lilly and SK Biopharmaceuticals.

It sees it as a way to reduce uncertainty in one of the most expensive, failure-prone industries on Earth.

Why Oncology Needs AI More Than Almost Any Other Field

Cancer is uniquely suited for AI because it is data-rich, heterogeneous, and dynamic.

Each patient’s disease can be measured and followed through: DNA, RNA, proteins, pathology, radiology, clinical records, treatment history, real-world outcomes/information.

But the human brain was not built to integrate all of that in real time.

That does not make clinicians obsolete. It makes multidisciplinary interpretation more important.

The future is not AI replacing oncologists.

The future is AI helping molecular tumor boards ask better questions

That is where precision oncology becomes operational.

The Missing Bridge: From Data to Decision

There is a dangerous illusion in healthcare:

That more data automatically means better medicine. It does not.

More data without interpretation creates confusion.

A 500-gene panel without clinical context can overwhelm physicians. A shorter focused panel may miss relevant alterations and cannot fully evaluate tumor-agnostic biomarkers.

RNA expression without biological interpretation becomes noise.

AI without quality data can become a very confident mistake.

The true opportunity is not just generating more molecular information.

It is building the bridge between all these:

- Tumor profiling

- Clinical records

- Bioinformatics

- AI-supported interpretation

- Molecular tumor boards

- Treatment decisions

That bridge is where the future of cancer care will be built.

And that is precisely where companies like Theranomics continue evolving — not only as providers of molecular testing, but as builders of an integrated precision oncology intelligence layer for hospitals, physicians, insurers, and patients.

The Payer and Hospital Case

For hospitals, AI-supported biomarker discovery can strengthen next-generation oncology programs.

It can help identify clinical trial candidates, support complex tumor board decisions, improve treatment sequencing, and differentiate institutions that want to offer true precision oncology rather than isolated molecular reports.

For insurers, the value is just as important.

Cancer therapies are becoming more expensive.

But expensive does not always mean appropriate.

AI-supported multi-omic interpretation could help justify high-cost therapies when the biology supports them — and avoid them when the probability of benefit is low.

That matters because the future of oncology reimbursement will not be based only on drug availability. It will increasingly depend on biological rationale.

How do we translate complex cancer biology into actionable, affordable, clinically useful decisions for real patients?

The Reality Check

AI will not magically cure cancer.

It will not replace clinical trials.

It will not compensate for poor samples, incomplete reports, biased datasets, or weak clinical interpretation.

And it should never be treated as an oracle.

AI is only as good as the biology it learns from, the datasets used to train it, and the humans who validate its conclusions. Garbage in, garbage out remains the first law of machine learning.

The challenge now is not only building better models, but validating them prospectively, across diverse populations, and proving that they improve real clinical outcomes.

That is the standard that matters.

Not hype.

Impact.

From Designing Drugs to Designing Systems

The next revolution in oncology will not come from one technology alone.

Not AI alone.

Not genomics alone.

Not CRISPR alone.

It will come from convergence.

AI helps us detect patterns.

Tumor molecular profiling and multi-omics tells us what is happening inside the tumor. Synthetic biology may allow us to design therapies that respond to those signals.

This is where cancer care begins to look less like static medicine and more like adaptive engineering.

A system that learns, predicts, designs and updates as the tumor evolves.

What Comes Next: The Digital Twin of Cancer

If AI can help us discover biomarkers and design better drugs…

What happens when we use all this information to build a virtual version of a patient’s tumor?

A digital mirror capable of simulating how cancer may evolve, how it may resist therapy, and which treatment strategy may work before we test it in the real patient.

That is where the next article will take us:

Digital twins in oncology.

Because the future of cancer care may not be only about choosing the next drug.

It may be about testing the next move before the tumor makes it.