There is a deceptively simple idea at the heart of modern oncology: treat the tumour's biology, not just its location. The tools that make this possible are biomarkers. They sit behind nearly every meaningful advance in cancer medicine over the past two decades, from the targeted therapies that transformed outcomes in breast and lung cancer, to the immunotherapies that now offer durable remissions across a dozen tumour types. And yet for many people, even those working in healthcare, the term remains abstract.
This post is an attempt to make it concrete. What biomarkers are, how they are classified, why they matter in oncology, and where the science is heading.
What is a biomarker?
The word sounds technical, but the underlying concept is not complicated. A biomarker is any measurable biological signal that provides useful information about a disease, its likely course, or its likely response to treatment. It is, in the words of the field, a decision-making tool.
That definition is deliberately broad. A biomarker can be a gene sequence, a protein measured in blood, a pattern visible under the microscope, or even an image. What makes something a biomarker is not what it is, but what it tells us and how reliably it tells it.
In drug development, biomarkers have proven to be transformative not just clinically but commercially. Data from the pharmaceutical industry shows that oncology trials incorporating biomarker strategies have dramatically higher success rates than those without them. Phase I to FDA approval success rates in oncology run roughly 6.7 percent overall, but with biomarker-selected populations, the progression rates through each clinical phase improve substantially. This is not coincidental. When you select the right patient for the right drug at the start, the signal becomes cleaner, the trial becomes smaller, and the outcome becomes more meaningful.
A framework for thinking about biomarker types
Biomarkers in oncology come in several forms, and understanding the differences between them is important for understanding how they are used in practice.
Molecular biomarkers: genes and proteins
These are the category most people associate with precision medicine, and for good reason. Molecular biomarkers include mutations, gene amplifications, gene fusions, protein overexpression, and gene expression signatures. They are measured from tumour tissue, blood, or increasingly from other bodily fluids.
At the genetic level, cancer driver mutations are of particular interest. These are alterations that actively promote tumour growth by constitutively activating signalling pathways that control cell survival and proliferation. Because these mutations are functionally dominant and often unique to the tumour, they are both prime therapeutic targets and excellent predictive biomarkers. Examples include EGFR mutations in lung cancer, KRAS and BRAF mutations in colorectal cancer, BRCA1 and BRCA2 mutations in breast and ovarian cancer, and HER2 amplification across multiple tumour types.
Beyond single gene mutations, molecular biomarkers also include broader genomic signals. Microsatellite instability (MSI) and mismatch repair deficiency (dMMR) reflect a breakdown in the cellular machinery that corrects replication errors, leading to an accumulation of mutations and heightened immune visibility. Tumour mutational burden (TMB), which quantifies the total number of somatic mutations per megabase of tumour DNA, is another genomic signal being used to predict immunotherapy response.
Histological and morphological biomarkers
Not all biomarkers require molecular analysis. Pathologists have long extracted prognostic information from the appearance of tumours under the microscope, and some of these morphological observations have been formalised as reproducible biomarkers.
A particularly well-established example is tumour budding in colorectal cancer. Tumour budding refers to isolated single cells or clusters of up to four cells found at the invasive margin of a tumour. It reflects the leading edge of tumour invasion and correlates strongly with lymphovascular invasion, lymph node metastasis, and worse outcomes. In endoscopically resected early-stage colorectal cancer, the presence of tumour budding identifies patients at higher risk of lymph node metastasis who may benefit from surgical resection. In stage II disease, it flags those who may need adjuvant therapy.
The International Tumour Budding Consensus Conference, held in Bern in 2016, produced standardised scoring criteria to make this assessment more reproducible across laboratories. This is a useful illustration of how a morphological observation, known for decades, becomes a validated biomarker through standardisation, reproducibility, and evidence-based consensus.
Tumour-infiltrating lymphocytes (TILs), the density and distribution of immune cells within and around a tumour, represent another histological biomarker gaining clinical traction, particularly in triple-negative breast cancer and other immunologically active tumour types.
Imaging biomarkers
Imaging biomarkers capture information about tumour size, structure, metabolic activity, and treatment response through imaging modalities. These include standardised measurements from CT, MRI, and PET scans. The RECIST criteria (Response Evaluation Criteria in Solid Tumours), used universally in clinical trials, are essentially a formalised imaging biomarker framework for assessing whether tumours are responding to treatment or progressing.
Quantitative imaging biomarkers are a growing area of research, with radiomics approaches using computational analysis of imaging features to extract prognostic information that is not visible to the naked eye.
Prognostic vs. predictive: a distinction that matters
One of the most important distinctions in clinical biomarker practice is between prognostic and predictive markers, and the two are frequently confused.
A prognostic biomarker tells you about the likely course of the disease independent of any specific treatment. It answers the question: what is likely to happen to this patient? Increasing PSA in prostate cancer predicts progression. BRCA1 or BRCA2 mutations in breast cancer patients indicate a higher likelihood of a second primary cancer. TP53 mutations in chronic lymphocytic leukaemia are associated with shorter survival.
A predictive biomarker tells you whether a specific treatment is likely to work. It answers the question: will this patient benefit from this intervention? HER2 amplification predicts response to trastuzumab. BRCA1 and BRCA2 mutations predict sensitivity to PARP inhibitors. PD-L1 expression on immune cells predicts response to atezolizumab in triple-negative breast cancer.
Some biomarkers do both. BRCA mutations are prognostic in that they carry information about disease trajectory, and predictive in that they identify candidates for PARP inhibitor therapy. Understanding which role a biomarker plays in a given clinical context is essential for using it correctly.
The role of biomarkers in oncology
Diagnosis and classification
Biomarkers allow pathologists to go beyond morphology in classifying tumours. Immunohistochemical markers are routinely used to identify tumour origin when metastases are of uncertain primary. CDX2 points to a gastrointestinal origin. TTF1 suggests lung or thyroid. GATA3 is associated with breast and urothelial origin. PAX8 indicates tumours from the urogenital tract. KIT and DOG1 mark gastrointestinal stromal tumours.
This diagnostic precision matters not just academically but therapeutically. A patient with a metastasis of uncertain origin can only receive targeted therapy if the tumour's molecular identity is established.
Treatment selection and targeted therapy
The most transformative application of biomarkers in oncology has been in guiding targeted therapy. When a drug is designed to inhibit a specific molecular target, the presence of that target in a patient's tumour is the most powerful predictor of whether the drug will work.
In non-small cell lung cancer, EGFR mutations, ALK rearrangements, ROS1 fusions, BRAF V600E mutations, MET exon 14 skipping alterations, and RET fusions are all now routinely tested because each predicts response to a specific approved agent. The treatment algorithm for metastatic lung cancer has gone from histology and stage to a detailed molecular decision tree, with biomarker status determining first-line therapy.
In colorectal cancer, RAS and BRAF mutation status determines eligibility for anti-EGFR therapy. Patients with KRAS or NRAS mutations do not benefit from cetuximab or panitumumab. BRAF V600E mutations identify a poor-prognosis subgroup now targetable with a combination of BRAF and MEK inhibition. HER2 amplification in colorectal cancer has emerged as another actionable target.
Immunotherapy and the new predictive landscape
Immunotherapy has introduced a new category of predictive biomarkers: those that predict sensitivity to immune activation rather than direct molecular inhibition.
MSI-H and dMMR status represent the most clinically established immunotherapy biomarkers. Tumours with deficient mismatch repair accumulate mutations rapidly, producing a high density of neoantigens that stimulate immune recognition. Pembrolizumab received full FDA approval for adult and paediatric patients with unresectable or metastatic MSI-H or dMMR solid tumours, achieving an objective response rate of 33.3 percent across multiple tumour types, with 77 percent of responders maintaining their response for at least twelve months.
The landmark 2017 FDA approval of pembrolizumab for MSI-H and dMMR solid tumours formalised the concept of biomarker-driven tumour-agnostic therapy, in which treatment approval is based on a shared molecular characteristic rather than the anatomical location of the tumour. This was a conceptual shift with enormous implications. For the first time, a drug was approved for patients based on what their tumours look like molecularly, not where in the body they arose.
TMB, which counts the total mutational load per megabase, is a related but distinct signal. Higher TMB indicates more neoantigens and a potentially more immunogenic tumour. The FDA approved pembrolizumab for TMB-high solid tumours in 2020. The use of TMB as a predictor of immunotherapy response has proven complex in certain tumour types, and challenges remain in establishing a universal pan-cancer threshold, particularly given the relationship between MSI-H status and mutational burden.
Synthetic lethality: a third class of predictive biomarker
Beyond direct target identification and immune response prediction, a third class of predictive biomarker has emerged based on the concept of synthetic lethality. Two genes are said to be synthetically lethal when loss of either alone is compatible with cell survival, but simultaneous loss of both causes cell death.
The most clinically successful application of this principle involves PARP inhibitors in tumours with BRCA1 or BRCA2 mutations. Cells with BRCA mutations cannot repair double-strand DNA breaks through homologous recombination. PARP inhibition blocks a compensatory single-strand repair pathway. The tumour cell, deprived of both repair mechanisms simultaneously, cannot survive. Normal cells, which retain functional BRCA genes, can tolerate PARP inhibition.
This mechanism has been extended beyond germline BRCA mutations to the broader concept of "BRCAness," encompassing tumours with alterations in other homologous recombination genes such as ATM, PALB2, RAD51, CHEK1, and CHEK2. Current research is mapping the predictive relevance of this broader mutational landscape to identify which patients with homologous recombination deficiency, beyond classic BRCA carriers, will respond to PARP inhibition or platinum chemotherapy.
Biomarkers across cancer types
Lung cancer
Lung cancer has become the paradigmatic example of molecularly guided oncology. The EGFR, ALK, and ROS1 testing that transformed outcomes in adenocarcinoma over the past fifteen years has been joined by a growing panel of additional targets. KRAS G12C mutations, long considered undruggable, are now targetable with sotorasib and adagrasib. MET amplification and RET fusions each have approved targeted agents. PD-L1 expression guides immunotherapy decisions in patients without actionable driver mutations.
Liquid biopsy has grown in significance for lung cancer, where tissue biopsy remains the gold standard for molecular profiling but its invasiveness and inability to provide real-time monitoring have driven adoption of blood-based testing to detect actionable alterations, monitor treatment response, and identify acquired resistance mechanisms.
Colorectal cancer
In colorectal cancer, MSI and MMR status serve dual roles: prognostic in early-stage disease and predictive for immunotherapy in metastatic disease. RAS and BRAF mutation testing are standard before initiating targeted therapy. Tumour budding, as discussed above, is now an established histological prognostic factor incorporated into pathology reporting frameworks. Molecular markers including RAS and NRAS mutations, BRAF mutations, HER2 amplification, microsatellite instability status, and tumour mutational burden have become central to therapy selection and prognosis in metastatic colorectal cancer.
The practical challenge is that these tests are not always ordered or interpreted consistently across institutions. Lynch syndrome, caused by germline mutations in MMR genes, has population-level implications because affected individuals are at elevated risk not just of colorectal cancer but of endometrial, ovarian, gastric, and other cancers. Universal MMR testing of colorectal cancer specimens has been recommended in many guidelines precisely because identifying MSI-H tumours also identifies families at risk.
Prostate cancer
In prostate cancer, PSA has been a clinical standard for decades as both a screening and monitoring tool, though its limitations as a screening biomarker are well documented. More recently, germline and somatic BRCA1 and BRCA2 mutations have emerged as actionable biomarkers, with PARP inhibitors olaparib and rucaparib approved for patients with BRCA-mutated metastatic castration-resistant prostate cancer. Genomic profiling of prostate tumours is increasingly standard in metastatic disease. AR-V7, a splice variant of the androgen receptor, is a predictive biomarker of resistance to enzalutamide and abiraterone, guiding treatment sequencing decisions.
The tumour-agnostic revolution
Perhaps the most consequential recent development in biomarker-guided oncology is the emergence of tumour-agnostic approvals. There are currently multiple tumour-agnostic approvals granted by the FDA, including immune checkpoint inhibitors and targeted therapy agents. Pembrolizumab was the first approved tumour-agnostic treatment for MSI-H and dMMR solid tumours in 2017, followed by approvals for TMB-high tumours, NTRK fusion-positive cancers with larotrectinib and entrectinib, BRAF V600E mutations with dabrafenib and trametinib, and RET fusions with selpercatinib.
In 2024, trastuzumab deruxtecan achieved tumour-agnostic approval for HER2-positive tumours with IHC 3+ status across multiple histologies, with the DESTINY-PanTumor02 study reporting an overall response rate of 37 percent across diverse cancer types.
The tumour-agnostic framework represents a philosophical shift. Treatment eligibility is no longer defined by where in the body a tumour originated but by what it looks like at a molecular level. A patient with a RET fusion in their lung cancer, thyroid cancer, or pancreatic cancer is, in important respects, the same patient from the perspective of targeted therapy.
Looking ahead: where biomarker science is going
Several developments are reshaping the biomarker landscape and will define oncology practice in the years ahead.
Liquid biopsy is moving from research tool to routine clinical utility. By detecting circulating tumour DNA, circulating tumour cells, and other tumour-derived material in blood, liquid biopsy offers the possibility of non-invasive molecular profiling, real-time monitoring of treatment response, early detection of resistance, and potentially earlier diagnosis of recurrence than imaging alone. Liquid biopsy has emerged as a valuable tool in precision medicine across lung, colorectal, breast, and prostate cancer, though standardisation of protocols, sample handling, and ensuring biomarker stability remain challenges for broader integration into routine clinical practice.
Next-generation sequencing has made comprehensive genomic profiling of tumours feasible at scale. Rather than testing one biomarker at a time, NGS panels assess hundreds of genes simultaneously, identifying actionable alterations that would be missed by sequential single-gene testing. Both ASCO and ESMO have issued guidance supporting comprehensive genomic profiling for patients with metastatic cancer where actionable alterations are plausible.
Artificial intelligence applied to pathology and radiology images is beginning to extract prognostic and predictive information from standard images that human review cannot reliably detect. AI-based analysis of haematoxylin and eosin stained slides has already demonstrated the ability to predict MSI status, BRCA mutation, and other molecular features without additional molecular testing in research settings.
Multi-omics integration, combining genomic, transcriptomic, proteomic, and epigenomic data from the same tumour, promises a more complete picture of tumour biology than any single layer of data can provide. The challenge is translating this complexity into clinically actionable outputs.
What this means for practice
The accumulation of validated biomarkers has changed the practical reality of oncology in ways that are easy to understate. A patient with metastatic non-small cell lung cancer in 2010 had essentially one treatment pathway: chemotherapy. Today that patient undergoes a panel of molecular tests, and the results may indicate any one of a dozen targeted or immunotherapy strategies, each with very different response rates, toxicity profiles, and implications for sequencing.
This creates new demands on every part of the system. Pathologists must produce accurate, reproducible biomarker results under time pressure. Oncologists must integrate an expanding body of evidence about which biomarkers matter in which contexts. Payers and health systems must evaluate which tests are clinically actionable and cost-effective. And patients must navigate a more complex landscape of results and decisions.
Getting biomarker science right, both the technical execution and the clinical interpretation, is not a niche interest for specialists. It is the foundation on which modern oncology is built.
Conclusion
Biomarkers are not an add-on to cancer care. They are increasingly its organising principle. The movement from treating cancer by site and stage to treating it by molecular identity is still underway, but it has already transformed outcomes for millions of patients. Every validated biomarker represents a more precise answer to the question that matters most in oncology: what will work for this patient?
The research horizon is genuinely exciting. Liquid biopsy, AI-assisted pathology, tumour-agnostic approvals, and multi-omics profiling are each expanding what is measurable and what is actionable. The challenge, as always, is translating discovery into validated, accessible, equitably deployed clinical tools. That translation, from biological observation to standardised test to changed clinical practice, is where the real work happens.
This post draws on course material from the CAS in Personalized Molecular Oncology at the University of Basel and is supplemented by current published evidence. It is intended as an educational overview for a general professional audience, not as clinical guidance.