Abstract

The rapid expansion of molecular biomarkers has fundamentally reshaped contemporary oncology practice, transitioning it from morphology-based classification toward biologically driven precision medicine. In line with current scientific advancements, oncology treatment practices will increasingly be patient-centered, driven by genetic analysis and targeted therapy. Personalized treatment based on next-generation sequencing (NGS) will assume a central role. A narrative review of recent literature was conducted, focusing on NGS, liquid biopsy and circulating tumor DNA (ctDNA), minimal residual disease (MRD), immuno-oncology biomarkers, and organ-specific clinical implementation. Biomarker-driven oncology now informs systemic therapy selection, adjuvant strategies, recurrence risk stratification, and multidisciplinary sequencing across major solid tumors, including lung, colorectal, breast, gastric, hepatobiliary, and prostate cancers. ctDNA-based MRD detection has emerged as a dynamic risk stratification tool, while immuno-oncology biomarkers such as programmed death-ligand 1, tumor mutational burden, and microsatellite instability guide checkpoint inhibitor therapy; these biomarkers have varying predictive performance. Future oncology practice will rely on integrated biomarker ecosystems combining genomics, dynamic monitoring, immunologic profiling, and artificial intelligence-assisted decision support systems.

Introduction

Oncology is undergoing a structural reorganization of its decision-making language to treat the diseas faced. Historically, tumor histology and anatomic stage were the primary coordinates for prognosis and treatment selection. On the other hand, in the molecular biomarker era, these coordinates remain essential, but they are increasingly complemented and sometimes superseded by biological features that predict therapeutic vulnerability, resistance, and recurrence risk (1-4) as well. This transition is not merely technological; it changes how clinicians conceptualize risk. A “high-risk” tumor is no longer defined only by size, nodal status, or grade, but in the age modern treatment era also by driver alterations, immune contexture, and the presence of molecular residual disease after definitive therapy (Figure 1) (5-11).

From a medical oncology daily perspective, biomarkers are matter for three practical reasons. First, they compress uncertainty: a predictive biomarker increases the probability that a treatment works for a given patient while reducing exposure to ineffective therapy and to avoid toxicity. Second, biomarkers can convert time into an actionable resource: longitudinal measurements [e.g., circulating tumor DNA (ctDNA)] allow clinicians to detect disease relapse earlier or to show resistance mechanisms without re-biopsy. Third, biomarkers reshape multidisciplinary care by influencing neoadjuvant sequencing, adjuvant intensification or de-escalation, and postoperative surveillance strategies areas where surgical oncology and medical oncology increasingly intersect together (5, 6, 9, 11, 12).

The last five years have been particularly dynamic in terms of oncology practice. Precision oncology reviews highlight a steady expansion of targeted therapies and a maturing ecosystem of trial designs (basket, umbrella, platform studies) that operationalize biomarker-driven hypotheses (2, 3). In conjuction with this, minimal residual disease (MRD) concepts have moved from hematologic malignancies into solid tumors, with ctDNA technologies enabling molecular detection below radiographic thresholds (5-11). Finally, artificial intelligence (AI) and multi-omics integration are shifting biomarker work from single assays to fused, multimodal predictors that may support clinical decision-making in clinical evaluation (13-15).

Biomarker Taxonomy and the Shift from Static to Dynamic Risk

As known that clinically, biomarkers are often categorized as diagnostic, prognostic, predictive, and monitoring biomarkers. In real daily practice these categories overlap, and the same biomarker can serve multiple roles. For example, microsatellite instability (MSI) can act as a prognostic marker in certain settings as in lung cancer, a predictive biomarker for immune checkpoint inhibitors as in colorectal cancer (CRC), and a clue to inherited cancer risk via mismatch repair deficiency pathways as in Lynch syndrome (16).

A useful contemporary distinction is between static biomarkers (typically measured once from baseline tissue) and dynamic biomarkers (measured repeatedly to capture evolution). Static biomarkers [e.g., epidermal growth factor receptor (EGFR) mutation in non-small cell lung cancer (NSCLC), human epidermal growth factor receptor 2 (HER2) amplification in breast or gastric cancer, breast cancer gene 2 alterations in prostate cancer] are foundational to targeted therapy selection and often guide first-line strategy (17-23). Dynamic biomarkers especially ctDNA capture changing tumor burden and emergent resistance, and are increasingly discussed as tools for MRD detection after curative-intent surgery or chemoradiation, respectively (5-11).

The dynamic view reframes postoperative surveillance: rather than waiting for radiographic recurrence, clinicians may stratify recurrence risk molecularly and tailor surveillance intensity and adjuvant therapy. However, earlier detection does not automatically imply improved outcomes; clinical utility ultimately depends on whether an intervention triggered by MRD status changes survival or quality of life, a point emphasized in emerging analyses of ctDNA-led recurrence detection (10, 11).

Enabling Technologies: Next-generation Sequencing (NGS), Liquid Biopsy, Digital Pathology, and Multi-omics NGS

NGS has become the main platform for identifying actionable alterations, characterizing resistance mechanisms, and enabling tumor-agnostic therapies in oncology practice. Contemporary reviews highlight the clinical translation of diverse biomarker classes oncogenic drivers, homologous recombination deficiency, gene fusions, and mutational signatures that often within panel based workflows that can be deployed across tumor types (1-4). The real-world challenge is less about the existence of sequencing and more about the operational pipeline: tissue adequacy, turnaround time, bioinformatics standardization, and how possible results are integrated into multidisciplinary decision pathways (18, 24).

Liquid Biopsy and ctDNA

Liquid biopsy encompasses multiple analytes: ctDNA, circulating tumor cells, extracellular vesicles, and other cell-free components. Among these, ctDNA has emerged as the most measurable clinical tool due to assay sensitivity, relative standardization, and direct linkage to tumor genomics (6-9). Narrative and systematic reviews describe applications in (I) baseline genotyping when tissue is unavailable, (II) monitoring treatment response and resistance, and (III) detecting MRD after curative-intent therapy (6-9,11).

The field distinguishes tumor-informed assays (custom panels based on the patient’s tumor) from tumor-agnostic assays (fixed panels without prior tumor sequencing). Tumor-informed approaches may offer higher specificity for MRD, while tumor-agnostic assays favor speed and simplicity; both approaches remain under active evaluation for clinical utility and cost-effectiveness as well (10, 11).

Digital Pathology and AI

Digital pathology is increasingly considered as a biomarker platform. AI models applied on hematoxylin and eosin images can predict genomic alterations, immune phenotypes, and even immunotherapy response in specific contexts, potentially reducing barriers where molecular testing is delayed or inaccessible (14, 15). Importantly, AI outputs should be treated as probabilistic decision aids rather than definitive biomarkers, and prospective validation plus interpretability remain prerequisites for clinical deployment (13-15).

Multi-omics and Multimodal Fusion

Multi-omics integration combines genomics, transcriptomics, epigenomics, proteomics, metabolomics, imaging, and clinical data to build more robust predictors than any single data layer. Recent reviews summarize AI-driven approaches in which including graph-based models and transformers for cross modal fusion and emphasize the need for standardization, reproducibility, and clinically meaningful endpoints (14, 15).

Organ-specific Translation: What Actually Changes in Clinical Practice (Figure 2, Table 1) (17, 20, 21, 24-26)

NSCLC

NSCLC represents the model of biomarker-driven systemic therapy. Contemporary guidelines and implementation reviews emphasize comprehensive molecular profiling for key drivers [EGFR, anaplastic lymphoma kinase, ROS proto-oncogene 1, B-Raf proto-oncogene (BRAF), MET exon 14 skipping, rearranged during transfection, neurotrophic tyrosine receptor kinase (NTRK), Kirsten rat sarcoma viral oncogene G12C mutation) because targeted therapy selection depends on accurate identification of actionable subsets and resistance pathways (18-20,25). Transactionally, delays and inconsistent testing remain major barriers; practical solutions include reflex testing, standardized pathways, and early integration of liquid biopsy when tissue is limited (18, 24).

In the perioperative setting, the biomarker story is expanding from driver genotyping to dynamic assessment of response and residual disease. ctDNA-based monitoring after curative-intent therapy has shown prognostic value in early-stage NSCLC and may enable risk adapted adjuvant strategies, although prospective interventional evidence is still emerging (7, 10). The clinical potential is clear: how to integrate molecular relapse signals without over-treatment or anxiety, and how to act when an MRD-positive result appears months before radiographic disease is detectable (5-11).

CRC

CRC biomarker practice involves both static and dynamic components. Static biomarkers include rat sarcoma virus and BRAF for targeted therapy selection and MSI status for immunotherapy eligibility and prognostic stratification (17). The most disruptive addition is ctDNA-based MRD. Multiple reviews synthesize evidence that postoperative ctDNA positivity is strongly associated with recurrence risk and may outperform clinicopathologic risk factors for MRD assessment (5, 6, 10, 11, 27, 28).

A key next-step question is whether MRD-guided treatment changes survival. Prospective work has begun to link MRD status to overall survival and to evaluate intervention strategies based on ctDNA results (11, 29). However, clinical utility requires more than prognostic correlation; pathways especially for adjuvant escalation, de-escalation, and the definition of actionable thresholds must be standardized (10, 11, 27-29).

Breast Cancer

Breast cancer remains structured by receptor biomarkers (estrogen receptor/progesterone receptor and HER2) as in luminal classification and genomic risk signatures, but molecular biomarker practice is evolving in two directions: resistance mapping in advanced disease and ctDNA applications in early-stage surveillance (20, 30-33). Estrogen receptor 1 mutations in ctDNA illustrate how liquid biopsy can operationalize endocrine resistance mechanisms and support treatment selection in hormone receptor-positive disease especially in progressive disease (30-32). Meanwhile, ctDNA MRD concepts in early-stage breast cancer are being actively investigated; reviews emphasize the consistent association between ctDNA positivity and recurrence risk while highlighting assay sensitivity, standardization, and interventional trial evidence as key gaps (33-35).

Gastric and Gastroesophageal Junction (GEJ) Cancers

In gastric/GEJ cancers, the biomarker landscape is expanding beyond HER2 and MSI to include Claudin 18.2 (CLDN18.2) as a therapeutic target. Recentlly, phase 3 trials have shown clinical benefit for zolbetuximab plus chemotherapy in CLDN18.2‑positive, HER2‑negative disease, motivating routine testing considerations (20-23). Reviews and prevalence studies suggest CLDN18.2 is common and may remain relatively stable over time, supporting its feasibility as a clinical biomarker (22, 23).

Hepatobiliary Cancers and Cholangiocarcinoma

Cholangiocarcinoma illustrates the value of biomarker stratification in rare, aggressive tumors. Fibroblast growth factor receptor 2 (FGFR2) rearrangements and isocitrate dehydrogenase 1 mutations are clinically actionable subsets. Targeted therapies (including FGFR inhibitors) have demonstrated activity in FGFR-altered disease, and multiple reviews address resistance mechanisms and safety profiles, emphasizing the importance of longitudinal molecular monitoring (25, 36, 37).

Prostate Cancer and Genitourinary Malignancies

Prostate cancer biomarkers increasingly span inherited and acquired alterations in DNA damage repair pathways, with poly (ADP-ribose) polymerase inhibitors demonstrating benefit in selected molecular subsets. Recent reviews summarize emerging biomarkers across genetic, RNA-based, metabolic, and epigenetic classes and discuss how molecular stratification may refine prognosis and therapy (26, 38). The near-term clinical direction is toward more systematic germline and somatic testing pipelines, earlier integration of targeted therapies, and biomarker guided combinations, while retaining careful toxicity management and validation (38).

Immuno-oncology Biomarkers: Promise, Friction, and Composite Approaches (Table 2) (12-16, 39)

The immunotherapy era introduced biomarkers that reflect tumor immune interaction rather than solely tumor-intrinsic genetics. PD‑L1 expression, tumor mutational burden (TMB), tumor-infiltrating lymphocytes, and MSI are the most widely discussed in oncology practice. Narrative reviews emphasize that predictive performance is inconsistent across tumor types and disease stages, limited by assay variability, sampling bias, and dynamic expression (12, 16, 39). MSI remains one of the most robust predictors of response to PD‑1 blockade in several contexts, and guideline recommendations emphasize standardized MSI testing approaches (17).

Beyond baseline prediction, biomarkers are increasingly used to understand resistance to therapy. In NSCLC, for example, reviews dissect tumor-intrinsic and microenvironmental drivers of immunotherapy resistance even in PD‑L1‑high disease, underscoring why single marker strategies often fail and motivating composite or longitudinal biomarker approaches (12).

Composite biomarker strategies combine tumor genomics (TMB/MSI), immunohistochemistry, spatial immune architecture, and dynamic signals such as ctDNA kinetics. This reflects a broader principle: immunotherapy response is a systems property, not a single gene event. Multimodal AI approaches that integrate histopathology, omics, and imaging may become particularly useful here, provided prospective validation demonstrates incremental value over existing clinical models (13-15).

MRD: From Prognostic Signal to Interventional Tool (Figure 3)

MRD in solid tumors is often termed molecular residual disease and refers to tumor derived fragments that persist after definitive therapy but remain below detection of conventional imaging. Reviews describe ctDNA MRD as a dynamic biomarker that can identify recurrence risk earlier than radiology and may support risk adapted adjuvant strategies (5, 6, 8-11).

CRC provides the most mature MRD evidence base, with prospective data linking MRD to outcomes and ongoing trials testing ctDNA-guided adjuvant therapy strategies (10, 11, 27-29). Breast and lung cancer are rapidly following, but the key obstacles are consistent: assay sensitivity (especially in low-shedding tumors), pre-analytic variables, false positives from clonal hematopoiesis, and the need for clinically actionable algorithms (8-11,33-35).

From a multidisciplinary viewpoint, MRD has surgical implications. It could influence adjuvant therapy decisions after R0 resection, surveillance intensity, and selection for clinical trials. Until now the field must avoid premature over reliance: MRD is a powerful risk stratifier, but patient benefit depends on evidence that MRD-guided actions improve survival, reduce toxicity, or meaningfully enhance quality of life (10, 11, 29).

Tumor-agnostic Biomarkers and the Expansion of “Biology-First” Treatment

Tumor-agnostic therapy targets are specific molecular alterations of whom independent of the tumor’s site of origin. Reviews highlight both the clinical promise and practical challenges: assay standardization, rare biomarker prevalence, equitable access to testing, and the need for multidisciplinary interpretation (40). NTRK fusions represent a canonical example of a biomarker enabling tumor-agnostic targeted therapy; clinical translation depends on robust fusion detection and careful diagnostic workflows (41).

Future Perspective

From Single Biomarkers to Biomarker Ecosystems (Table 3) (13-15)

The next phase of biomarker oncology is likely to be characterized by integration rather than proliferation. Instead of adding isolated markers, clinical value will come from (I) standardized pipelines, (II) longitudinal monitoring frameworks, and (III) multimodal fusion of omics, imaging, histopathology, and real-world clinical data for decision support (13-15).

AI-driven multi-omics integration reviews emphasize two themes. First, model performance must translate into clinical utility, meaning measurable improvement in decisions compared with current standards. Second, interpretability, robustness, and governance are not optional; they determine whether a model can be trusted in heterogeneous real-world settings (13-15).

Transactionally, the future is also about systems: turnaround time, reimbursement, laboratory capacity, and multidisciplinary tumor boards capable of interpreting complex results. Biomarker innovation will not achieve its potential unless health systems can deliver timely, standardized testing and embed results into care pathways.

Practical Recommendations for Clinical Integration

• Use guideline concordant comprehensive profiling in biomarker-driven diseases (e.g., NSCLC) and ensure reflex testing pathways in daily practice (17-19,24).

• Prefer validated, clinically actionable biomarkers; treat exploratory multi-omics predictors as decision aids until prospectively validated (13-15).

• For MRD/ctDNA, integrate results into predefined clinical pathways (adjuvant escalation/de-escalation, trial referral) and communicate uncertainty clearly (5-11,27-29,33-35).

• Anticipate tumor heterogeneity: when tissue is limited, consider liquid biopsy as complementary rather than substitutive when feasible (6-9,11).

• Maintain multidisciplinary review of complex findings, especially for tumor-agnostic indications and rare alterations (40, 41).