Precision Oncology: How Tumour Profiling Is Changing Cancer Treatment

Two Patients, One Diagnosis, Different Outcomes

Consider two people sitting in the same oncologist's waiting room on the same afternoon. Both have been told they have non-small cell lung cancer. Both are at stage III. Both are otherwise healthy adults in their mid-fifties. On paper, their prognoses look nearly identical. But one of them carries an activating mutation in the EGFR gene, while the other does not. That single molecular difference will determine whether a targeted oral tablet shrinks their tumour within weeks, or whether they proceed to a more aggressive and far less specific treatment regimen. Their stories will diverge not because of chance, but because of biology that oncologists can now actually read.

This is the central promise of precision oncology: that by understanding the precise genetic fingerprint of a tumour, clinicians can match individual patients to therapies that are far more likely to work, while sparing them the toxicity of treatments that would almost certainly fail. It is a shift in philosophy that has been building for decades, accelerating sharply since the sequencing of the human genome in 2003, and it is now reaching a level of clinical maturity that makes it relevant to anyone receiving a cancer diagnosis today. If you or someone you care about is navigating a cancer diagnosis, understanding what tumour profiling is and what it can tell your medical team is no longer a luxury: it is a practical necessity.

The field sits at the intersection of genomics, pharmacology, and data science. It draws on technologies that would have seemed extraordinary just twenty years ago, including next-generation sequencing capable of reading millions of DNA fragments simultaneously, liquid biopsies that detect tumour DNA floating free in the bloodstream, and artificial intelligence systems that synthesize genomic reports into clinically actionable guidance. Understanding how these tools work, what they look for, and what their limitations are puts you in a far stronger position when you are sitting across from your oncologist trying to make decisions under enormous pressure. This article walks through each layer of that understanding, from the biology of mutations to the algorithms now reading genomic reports, in plain language that does not require a medical degree.

What Tumour Profiling Involves

Tumour profiling begins with tissue. When a biopsy is performed, the removed sample contains both cancerous cells and healthy surrounding cells. Pathologists first confirm the diagnosis and classify the cancer type under a microscope, as they always have. But in a precision oncology workflow, the sample then travels to a molecular laboratory where its DNA, RNA, and sometimes protein expression patterns are analysed in extraordinary detail. The workhorse technology for this analysis is next-generation sequencing, commonly abbreviated to NGS.

Next-generation sequencing works by fragmenting the tumour's DNA into millions of short segments, sequencing all of them in parallel, and then computationally reassembling those fragments into a coherent picture of the tumour's genome. This is fundamentally different from older single-gene tests, which could only look at one mutation at a time. An NGS panel can simultaneously examine hundreds of genes for thousands of different mutations, copy number alterations, gene fusions, and markers of genomic instability. The resulting report is dense with data, running to dozens of pages in its raw form, and it requires considerable expertise to interpret meaningfully.

One of the most widely used comprehensive genomic profiling tests is FoundationOne CDx, developed by Foundation Medicine, a company now operating under the Roche umbrella. FoundationOne CDx analyses more than 300 cancer-related genes and is approved by the US Food and Drug Administration as a companion diagnostic for multiple targeted therapies across tumour types. When a patient's sample is sent to Foundation Medicine, the laboratory performs deep sequencing on the tissue, generates a report identifying all relevant genomic alterations, and provides a tumour mutational burden score, a microsatellite instability status, and a list of biomarker-matched therapy options. This kind of comprehensive profiling has become a standard recommendation for patients with advanced solid tumours, and it is increasingly being incorporated into treatment planning at earlier stages as well.

The concept of biomarkers is central to understanding what profiling is actually looking for. A biomarker, in this context, is any measurable molecular characteristic of a tumour that predicts how that tumour will behave or how it will respond to a given treatment. Some biomarkers are predictive: they tell you whether a specific drug is likely to work. Others are prognostic: they tell you something about how aggressive the cancer is likely to be regardless of treatment. The mutations that precision oncology has made most clinically actionable are predictive biomarkers, and a handful of them have transformed the treatment of specific cancer types in ways that would have seemed implausible to oncologists practicing even fifteen years ago. You can learn more about the broader framework in which these tools operate by reading our overview of what precision medicine actually means in clinical practice.

The Mutations That Change Everything

Not all mutations are created equal. A typical tumour carries hundreds or even thousands of somatic mutations: changes in the DNA that accumulated over time through replication errors, environmental exposures, or defective repair mechanisms. The vast majority of these are passenger mutations, random bystanders that do not meaningfully drive the cancer's growth. A smaller number are driver mutations, the ones that actively promote proliferation, invasion, or resistance to cell death. Precision oncology focuses almost entirely on driver mutations, particularly those in genes for which targeted therapies already exist or are in clinical development.

EGFR, the epidermal growth factor receptor gene, is one of the most extensively studied oncogenes in cancer biology. Activating mutations in EGFR are present in roughly 10 to 15 percent of non-small cell lung cancers in Western populations, and in 30 to 40 percent of cases in East Asian populations, according to data compiled by the International Association for the Study of Lung Cancer. These mutations keep the EGFR protein in a permanently switched-on state, continuously signaling cells to divide. For patients carrying these mutations, targeted therapies that inhibit EGFR have dramatically changed outcomes.

ALK gene rearrangements represent another landmark in this story. In 2007, researchers Soda and colleagues published a paper in Nature identifying EML4-ALK fusions in a subset of non-small cell lung cancer patients. This fusion creates an abnormal protein that drives uncontrolled growth, and it is present in approximately 3 to 5 percent of NSCLC cases. The identification of this driver mutation led directly to the development of a class of drugs called ALK inhibitors, which have produced response rates and survival outcomes that far exceed what standard chemotherapy could achieve in this population.

HER2, formally known as ERBB2, is a receptor tyrosine kinase that is amplified or overexpressed in approximately 15 to 20 percent of breast cancers, as well as in subsets of gastric, lung, and other cancers. HER2 amplification means there are too many copies of the gene, flooding the cell's surface with HER2 receptors and driving aggressive growth. This was one of the first oncogenic drivers to be successfully targeted, and the development of HER2-directed therapy transformed what had been one of the most lethal breast cancer subtypes into one with substantially improved survival rates.

BRCA1 and BRCA2 are tumour suppressor genes that encode proteins critical to repairing double-strand breaks in DNA through a process called homologous recombination. When one or both copies of these genes are mutated or silenced, cells lose this repair capability and become genomically unstable, accumulating mutations at an accelerated rate. BRCA mutations are most commonly associated with hereditary breast and ovarian cancer, but they are also found in prostate cancer, pancreatic cancer, and other tumour types. The clinical importance of BRCA status extends beyond prognosis: it has opened the door to a class of targeted agents called PARP inhibitors, which exploit the specific vulnerability that BRCA-deficient cells have to a synthetic lethal mechanism.

KRAS deserves special mention because it was, for many years, considered undruggable. KRAS is one of the most frequently mutated oncogenes in human cancer, particularly in pancreatic ductal adenocarcinoma, colorectal cancer, and lung adenocarcinoma. Its protein product acts as a molecular switch, and oncogenic mutations lock it in the permanently activated position. Despite decades of effort, the protein's structure made it extremely difficult to design small molecules that could inhibit it. That changed with the development of KRAS G12C-specific inhibitors, which exploit a unique cysteine residue present in one specific KRAS variant. The approval of sotorasib in 2021 marked the first time a KRAS inhibitor had reached patients after nearly four decades of failed attempts, making it one of the most celebrated advances in oncology drug development in recent memory.

PD-L1 occupies a different category from these classical oncogenic drivers. Rather than being a mutation that drives tumour growth, PD-L1 expression is a marker of how a tumour interacts with the immune system. Programmed death-ligand 1 is a protein that tumour cells express on their surface, and it binds to PD-1 receptors on T cells, effectively telling those immune cells to stand down. High PD-L1 expression is one indicator that a tumour may respond to checkpoint inhibitor immunotherapy, though it is neither necessary nor sufficient on its own.

Targeted Therapies: Matching Drug to Mutation

The whole point of identifying these mutations is to match them to treatments that specifically exploit them. This is the domain of targeted therapy, and the clinical results in well-matched populations have been transformative enough to permanently change how oncologists approach treatment selection.

Trastuzumab, sold under the brand name Herceptin, was one of the first molecularly targeted antibodies to reach widespread clinical use in oncology. Developed in the 1990s and approved by the FDA in 1998 for HER2-positive breast cancer, trastuzumab binds directly to the extracellular domain of the HER2 receptor, blocking its signaling activity and marking HER2-overexpressing cancer cells for destruction by the immune system. Before trastuzumab, HER2-positive breast cancer was one of the most aggressive subtypes, with high rates of early relapse and poor overall survival. After trastuzumab, and with the subsequent development of additional HER2-directed agents including pertuzumab and trastuzumab emtansine, outcomes for this subtype improved dramatically. This is a concrete demonstration of what happens when a biomarker and a matched therapy come together: a previously poor-prognosis cancer becomes manageable for many patients.

Osimertinib, marketed as Tagrisso by AstraZeneca, represents the third generation of EGFR-targeted therapy in non-small cell lung cancer. First-generation EGFR inhibitors like gefitinib and erlotinib were transformative when they were introduced, but most patients eventually developed resistance through a secondary mutation called T790M. Osimertinib was designed specifically to inhibit both the original activating mutations and T790M, and clinical trials have shown it to significantly extend progression-free survival in patients with EGFR-mutated NSCLC. The FLAURA trial, published in the New England Journal of Medicine in 2018, demonstrated that osimertinib in the first-line setting outperformed earlier EGFR inhibitors on multiple survival endpoints. The drug has now become the standard first-line choice for EGFR-mutated advanced NSCLC at most major cancer centres.

Olaparib, a PARP inhibitor developed by AstraZeneca and now approved across multiple tumour types including BRCA-mutated breast, ovarian, and prostate cancers, works through the concept of synthetic lethality. Cancer cells with defective BRCA-mediated homologous recombination already have a compromised DNA repair system. PARP is a different repair enzyme that handles a different type of DNA damage. When you inhibit PARP in a cell that already cannot perform homologous recombination, the cell accumulates irreparable DNA damage and dies. Crucially, normal cells with one functional copy of BRCA can survive PARP inhibition because they retain repair capacity: the toxicity is therefore relatively selective for the cancer cells. This elegance of mechanism has made PARP inhibitors one of the most scientifically compelling advances in oncology pharmacology in the past two decades.

Immunotherapy and the Role of PD-L1

Immunotherapy has added another dimension to precision oncology, one that does not rely on targeting a specific oncogenic driver but instead on releasing the brakes that tumours place on the immune system. The checkpoint inhibitors, particularly agents that block the PD-1/PD-L1 axis, have produced durable remissions in subsets of patients with cancers that were previously considered almost uniformly lethal at advanced stages, including melanoma, non-small cell lung cancer, and bladder cancer.

The key question for immunotherapy is predicting who will respond. PD-L1 expression, measured by immunohistochemistry on tumour tissue, was the first biomarker used in clinical practice for this purpose. Patients with high PD-L1 expression, typically defined as 50 percent or more of tumour cells staining positive, show higher response rates to PD-1 inhibitors like pembrolizumab in first-line lung cancer treatment. The KEYNOTE-024 trial, led by researchers including Martin Reck at LungenClinic Grosshansdorf, established pembrolizumab monotherapy as a first-line standard for high PD-L1 expressors, significantly changing the treatment paradigm for a substantial subset of lung cancer patients.

However, PD-L1 alone is an imperfect predictor. Some patients with low or absent PD-L1 expression respond to immunotherapy, while some with high expression do not. Tumour mutational burden, a measure of how many somatic mutations are present per megabase of coding sequence, has emerged as a complementary biomarker. The reasoning is straightforward: tumours with many mutations produce more abnormal proteins, called neoantigens, that the immune system can potentially recognise. Higher tumour mutational burden is associated with greater neoantigen load and, in some contexts, with higher response rates to checkpoint inhibitors. Microsatellite instability, which reflects a defective mismatch repair system and leads to an accumulation of small insertions and deletions throughout the genome, is another biomarker that predicts immunotherapy response across tumour types, and it earned the first tumour-agnostic FDA approval in cancer history when pembrolizumab was approved for MSI-high solid tumours in 2017.

Liquid Biopsy and ctDNA Monitoring

One of the most significant practical advances in tumour profiling over the past decade has been the development of liquid biopsy, a technology that can detect and analyse circulating tumour DNA, abbreviated ctDNA, from a simple blood draw. When cancer cells die, they release fragments of their DNA into the bloodstream. These fragments carry the same mutations as the primary tumour and any metastatic deposits, making them a molecular window into the cancer's current state without the need for an invasive tissue biopsy.

The clinical applications of liquid biopsy are expanding rapidly. In patients with advanced solid tumours, ctDNA analysis can identify actionable mutations when tissue biopsy is not feasible or when the available tissue is insufficient for comprehensive genomic profiling. This is particularly valuable in tumour types like advanced lung cancer, where re-biopsy after progression can be technically challenging and carries procedural risks. Companies like Guardant Health, with their Guardant360 platform, and Foundation Medicine have developed validated liquid biopsy tests that are now widely used in clinical practice.

Perhaps more exciting than its diagnostic applications is liquid biopsy's potential role in treatment monitoring. Because ctDNA levels in the blood reflect the overall tumour burden in the body, serial measurements can track how well a treatment is working in near real time. A falling ctDNA level after starting a targeted therapy is a strong early indicator that the treatment is effective, often weeks before changes are visible on imaging. Conversely, a rising ctDNA level can signal emerging resistance and allow clinicians to consider a treatment switch before overt clinical progression occurs. Researchers at institutions including the Institute of Cancer Research in London and Memorial Sloan Kettering Cancer Center have published extensively on the use of ctDNA dynamics as a pharmacodynamic marker during treatment, and prospective trials evaluating ctDNA-guided treatment decisions are ongoing across multiple tumour types.

How AI Reads Genomic Reports

A comprehensive genomic profiling report from a platform like FoundationOne CDx can identify dozens of genomic alterations simultaneously. For a busy oncologist seeing many patients, synthesising all of that information, cross-referencing it against a rapidly evolving evidence base of trials and approved therapies, and translating it into a personalised treatment plan is an enormous cognitive challenge. This is precisely where artificial intelligence is beginning to contribute meaningfully to precision oncology workflows.

AI systems applied to genomic report interpretation typically work in one of several ways. Some use curated knowledge bases that link specific genomic alterations to evidence-graded therapy options, clinical trial eligibility criteria, and prognostic information. Others apply machine learning models trained on large datasets of genomic profiles and matched clinical outcomes to generate predictions about likely treatment responses. The most sophisticated systems combine both approaches, layering statistical pattern recognition over structured clinical knowledge in ways that surface non-obvious associations a human reader might miss.

Research groups including the oncology informatics team at the Broad Institute and computational biologists at the Dana-Farber Cancer Institute have published work on machine learning approaches to predicting treatment response from multi-omic tumour profiles. The challenge is not just identifying which mutations are present but understanding how they interact: a tumour with an EGFR activating mutation alongside a concurrent MET amplification, for example, is far more likely to show primary resistance to EGFR inhibitors than one with the EGFR mutation alone. These co-occurring alterations and their interaction effects are the kind of complexity that well-designed AI systems can handle more reliably than unaided human interpretation across thousands of possible combinations. The broader potential of these computational approaches is something we also explore in our article on quantum computing applications in cancer therapy development.

Clinical decision support tools incorporating AI-driven genomic interpretation are now deployed at several major academic medical centres. The Molecular Tumor Board, a multidisciplinary panel including oncologists, molecular pathologists, bioinformaticians, and pharmacologists, has become a standard institutional mechanism for reviewing complex genomic cases. AI tools increasingly support these boards by pre-processing reports, flagging the most clinically relevant alterations, identifying applicable clinical trials, and generating structured summaries that allow the human experts to focus their discussion on the highest-priority decisions. This human-AI collaboration model reflects a maturing understanding that the goal is not to replace expert clinical judgment but to augment it with computational capabilities that no individual clinician can replicate alone.

Questions about who controls the resulting data, and what happens to your genomic information after it enters a laboratory and clinical information system, are ones that every patient undergoing profiling should consider carefully. Understanding your rights around genomic data is not a peripheral concern: it is central to informed consent in this era of molecular medicine. Our article on who owns your medical records and genomic data addresses these questions directly.

What to Ask Your Oncologist

If you have been diagnosed with cancer, particularly an advanced solid tumour, the question of whether comprehensive genomic profiling is appropriate for your situation is one you should raise actively with your oncologist rather than waiting to see if it is offered. Most major oncology guidelines, including those from the National Comprehensive Cancer Network and the European Society for Medical Oncology, now recommend genomic profiling for patients with advanced or metastatic solid tumours before starting systemic therapy, precisely because the results can fundamentally alter the treatment plan.

Specific questions worth raising in your consultation include: whether a comprehensive NGS panel has been ordered or is recommended for your tumour type, which specific biomarkers are being assessed and why, whether your case will be reviewed by a molecular tumour board, and whether there are any ongoing clinical trials for which your genomic profile might make you eligible. Clinical trials in precision oncology often enrol patients based on molecular eligibility criteria rather than tumour type, meaning a trial for a KRAS G12C inhibitor might accept patients with lung cancer, colorectal cancer, or any other solid tumour carrying that specific mutation. These basket trials have opened new avenues for patients with rare mutations who might otherwise have no targeted options.

It is also worth asking about the turnaround time for profiling results and how the results will be communicated to you. Comprehensive NGS testing typically takes two to three weeks from tissue receipt to report, and some patients find it frustrating to wait for results when they are anxious to begin treatment. Your oncologist should be able to explain whether it is appropriate to begin therapy before profiling results return, or whether the potential benefit of a matched targeted therapy justifies the wait. In many situations, particularly for EGFR and ALK testing in non-small cell lung cancer, waiting for results before initiating systemic therapy is the clearly recommended approach.

Insurance coverage for comprehensive genomic profiling varies by country, insurer, and clinical context. In the United States, Medicare covers FoundationOne CDx for advanced solid tumours under specific criteria, and many private insurers follow similar policies. If coverage is denied, most major profiling laboratories have financial assistance programs, and the out-of-pocket cost of comprehensive profiling is often modest compared to the potential cost of ineffective treatment. Your oncology team's patient navigator or financial counsellor can help you navigate coverage questions before the test is ordered.

Finally, it is worth understanding that tumour profiling is not a one-time event. Tumours evolve under treatment pressure, acquiring new mutations that can confer resistance to therapies that were initially effective. Serial profiling, whether through repeat tissue biopsy or liquid biopsy, can reveal the molecular basis of resistance and guide next-line treatment selection. The genomic landscape of your cancer at diagnosis may look quite different from what is driving it at progression, and staying engaged with the molecular dimension of your disease over the course of treatment is one of the most proactive things you can do. Precision oncology is not a passive process: it rewards patients who ask informed questions and remain active participants in their own care decisions.

The story of precision oncology is still being written. Researchers estimate that the number of actionable genomic alterations will continue to grow as more targets are validated, more targeted therapies receive approval, and the ability to combine targeted agents with immunotherapy and other modalities becomes better understood. The two patients in the waiting room at the beginning of this article represent not just a clinical reality but a direction of travel: toward a future in which every cancer patient's treatment is guided by the precise molecular characteristics of their specific disease rather than by the statistical average of a population. That future is not fully here yet, but it is far closer than most people realise, and the tools to begin moving toward it exist right now. For a broader understanding of how genomic data intersects with personalised treatment decisions across medicine, the framework of pharmacogenomics and how your genes affect drug response provides essential complementary context.