Open research questions in Cancer Genomics and Diagnostics

What the literature leaves open

• Future precision oncology systems will likely integrate: • AI-assisted ctDNA surveillance, • wearable biosensor systems, • real-time molecular monitoring, Page | 79 Artificial Intelligence, ctDNA, and Biomarker-Driven Precision Oncology: Emerging Molecular Strategies for Cancer Diagnosis…

  Artificial Intelligence, ctDNA, and Biomarker-Driven Precision Oncology: Emerging Molecular Strategies for Cancer Diagnosis and Therapeutic Response Prediction · 2026 · DOI

• However, the relationships among existing CNS frameworks and their predictability from molecular data other than whole-genome sequencing remain unclear.

  A Pan-Cancer Multi-Omic Analysis of Copy Number Signature Clusters and Genomic Instability · 2026 · DOI

• of future classification. 1.4 Research Questions 1. What are the principal multi-omics strategies in tumor classification? approaches 2. What is the added value of integrating multitumor omics classification and diagnostic accuracy? 3. What are the obstacles and opportunities for using multi-omics technologies in oncology?

  EMERGING ROLE OF MULTI-OMICS APPROACHES IN TUMOR CLASSIFICATION · 2026 · DOI

• ABSTRACT Precision oncology relies primarily on DNA‐level alterations for therapeutic decisions, but the extent to which driver mutations propagate to protein abundance has not been systematically evaluated.

  Pan‐Cancer Quantification of Driver Alteration Transmission Across Molecular Layers Reveals Limited Propagation to Protein Abundance · 2026 · DOI

• The long-term solution lies in establishing regional molecular diagnostic capacity, whether through in-house platform develop- ment, regional laboratory networks, or participation in structured twinning programs with internationally accredited centers (29). The MNP-Outreach Consortium, for example, was specifically estab- lished to facilitate global adoption of methylation-based CNS tumor classification in LMICs and represents the type of collaborative infrastructure that institutions such as KHCC should actively engage with (30).

  Access to molecular diagnostics for CNS tumors through international outsourcing: experience from Jordan · 2026 · DOI

• Conventional surveillance tools, including imaging, lactate dehydrogenase (LDH), and S100B, are limited by their sensitivity and specificity, particularly in patients with low tumor burden.

  Circulating Tumor DNA as a Biomarker for Melanoma Prognosis and Therapy · 2026 · DOI

• Both strat- egies suffer from reduced sensitivity: the former fails to detect rare or unique patient-specific mutations, which make up the majority of somatic variants [48, 49], while the latter is limited by the inefficiency of duplex consen- sus collapsing, which yields a limited number of usable ctDNA fragments [50, 51].

  Tumor-naïve ctDNA detection with deep learning-enhanced error suppression for sensitive mutation calling · 2026 · DOI

• While patient-level copy number alteration (CNA) differences have been investigated extensively in large cohorts, their intratumoral heterogeneity remains understudied.

  A pan-cancer single-cell analysis of intratumoral copy number diversity and evolution · 2026 · DOI

• Fraction genome altered (FGA), a measure of chromosomal instability derived from next-generation sequencing (NGS), is already used in clinical practice but its prognostic value in early-stage NSCLC is not well established.

  Fraction genome altered to predict survival in early-stage non–small cell lung cancer. · 2026 · DOI

• The future of LRS in cancer liquid biopsy will depend on continued advances in sequencing chemistry, computational analysis, and assay design. Emerging studies suggest that machine learning and artificial intelligence (AI) tools may improve basecalling accuracy, multimodal integration, and longitudinal interpretation of cfDNA (70–72). However, while these advances are promising, the most likely role for LRS may be as a complement, rather than a replace- ment, to short-read assays. The strengths of LRS platforms may be integrated through hybrid sequencing strategies. These approaches leverage pairing long-range genomic content from LRS with high per- base accuracy of short-read NGS, resulting in more precise tumor profiling (73, 74). Such hybrid workflows have also been proposed as strategies to mitigate high costs and performance limitations associ- ated with LRS alone (75–77).

  Long-read sequencing for cancer liquid biopsy: advancing precision oncology · 2026 · DOI

• This section describes a series of research challenges related to PMT, derived from the gaps identified in the reviewed studies. Feature engineering Feature engineering is a key dimension in PMT techniques and an area that deserves further investigation. Many studies, particularly within the AL channel, rely on broadly similar sets of features, with static features often constrained by the capabilities of the feature extraction tools used. While these features are commonly used, it remains an open question whether additional or more specialized characteristics could contribute to capturing mutant killability more accurately. Beyond feature selection, the way different sources of information are combined also 1 3Automated Software Engineering (2026) 33:84 Page 37 of 43 84 deserves further attention. Apart from the hybrid approach proposed by Xu et al. (2023), the exploration of approaches that integrate features from multiple perspectives has received limited attention. Likewise, ensemble strategies have primarily focused on combining different ML algorithms (Aghamohammadi and Mirian-Hosseinabadi 2021), rather than aggregating models built on different feature representations or prediction paradigms. Finally, prior work has included ablation studies and explainable AI techniques to analyze feature relevance, especially in AL-based approaches (Guerrero-Contreras et al. 2025). However, more recent PMT techniques —such as those based on deep neural models or contrastive learning—– introduce increasingly complex representations, making interpretation more challenging. This raises open questions regarding how to effectively analyze and understand which aspects of the input data contribute most to modeling mutant killability under these newer paradigms. Programming languages As shown in Section 3, nearly all existing PMT studies have focused exclusively on Java programs. Currently, it remains unclear whether the effectiveness of the most recent approaches —NL channel and MSL— generalizes to other languages, especially those that differ significantly from Java in structure and paradigm. These differences are not limited to syntax. Programming languages can also influence the types of mutations applied and their relevance, how tests are written and executed, and how code is documented or named. Evaluating PMT in a broader range of languages would help assess its generalizability and identify potential language-specific challenges. Mutation tool and mutation operators The current body of PMT research has analyzed method-level operators provided by established tools such as PIT or Major. However, the applicability of PMT to other categories, such as class-level or performance-oriented operators (Delgado-Pérez et al. 2021; Wu et al.

  Methodological pitfalls in predictive mutation testing: threats, impact and open challenges · 2026 · DOI

• The detection and characterization of pathological alterations in the length of circular DNA represent a critical frontier in molecular diagnostics and disease mechanism research. As detailed in this review, the methodological arsenal available to researchers and clinicians is both extensive and evolving, ranging from foundational biochemical techniques to cutting-edge sequencing and imaging technologies. The optimal approach is never universal but is meticulously dictated by the specific biological question—be it screening for a common mitochondrial deletion, validating a novel oncogenic ecDNA structure, or discovering the complete circular genome landscape of a tumor. Key decision factors include the required resolution (from kilobase-scale changes to single-nucleotide breakpoints), necessary throughput, demand for absolute quantification, and available resources. The strategic integration of complementary methods—such as using long-read sequencing for discovery followed by ddPCR for ultrasensitive monitoring—often provides the most robust and clinically actionable insights. Looking forward, several promising directions will shape the next generation of circular DNA analysis, moving beyond mere detection towards a systems-level understanding of its functional impact. 1. Integration of Multi-Omics Data. The future lies in correlating the physical structure and length of circular DNA with its functional outputs. Simply knowing a circle is elongated is insufficient; understanding its transcriptional and epigenetic activity is paramount. Future workflows will integrate long-read sequencing (for length and sequence) with companion assays like RNA-seq from © Under CC BY-NC-ND 4.0 International License | Annals of Rejuvenation Science 1(2) 16 the same sample to link specific ecDNA structures to massive oncogene overexpression (Hung et al., 2021). Furthermore, leveraging the native epigenomic detection capability of Nanopore sequencing will allow researchers to simultaneously map the methylation landscape of a circular molecule, providing clues about its origin and regulatory state (Liu et al., 2021). This multi-modal integration will be essential for distinguishing passenger from driver alterations and for understanding how circular DNA structure directly influences gene regulation. 2. Advancement of Bioinformatics Tools. The power of long-read sequencing is currently bottlenecked by bioinformatic challenges. There is a pressing need for more sophisticated, dedicated algorithms for the de novo assembly, phasing, and structural variant calling specifically from circular DNA sequences.

  The Circular DNA Size Code · 2026 · DOI

• Conflict of interest The deep multimodal data fusion of PET/MRI and liquid biopsy is driving the evolution of cancer diagnosis and treatment toward a multimodal approach. This strategy synergizes macroscopic imag- ing information with microscopic molecular data, demonstrating clear value in early tumor detection, heterogeneity analysis, dy- namic treatment monitoring, and precise prognostic stratification. However, its clinical translation still faces core challenges, including a lack of standardization, algorithmic bottlenecks, and insufficient high-level evidence. Moving forward, leveraging artificial intelli- gence and multi-omics technologies to build standardized data analysis platforms and validate clinical utility through prospective trials will be essential. The advancement of this integrated paradigm will provide critical technical support for the transition from population-based treatments to individualized precision medicine, ultimately enhancing the systematic and effective management of cancer. Future directions include: developing AI-driven, multimodal data fusion platforms to achieve end-to-end optimization from raw data to clinical decision-making; exploring the integration of multi-dimensional liquid biopsies beyond blood (such as cerebro- spinal fluid and urine) with site-specific imaging for specialized types like central nervous system tumors; building personalized dynamic monitoring networks based on regular liquid biopsies and key time-point PET/MRI scans to enable predictive healthcare; and ultimately forming a closed-loop, integrated diagnostic and therapeutic system. For instance, by leveraging The author(s) declared that this work was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

  Multimodal data fusion: integrating PET/MRI and liquid biopsy for a holistic view of cancer biology · 2026 · DOI

• Insensitive to micrometastases; high cost; radiation exposure Early detection, MRD monitoring, clonal evolution tracking, resistance mutation detection Lacks spatial context; low sensitivity in early-stage cancers; pre-analytical variability; lack of standardization molecular “ground truth” (e.g., specific driver mutations, methyla- tion status). This allows for the development of more robust and biologically plausible fusion models that explicitly link the imaging phenotype with the molecular genotype, moving beyond prediction to a deeper biological understanding and enhancing the reliability of the model.

  Multimodal data fusion: integrating PET/MRI and liquid biopsy for a holistic view of cancer biology · 2026 · DOI

• PET/MRI: an integrated imaging platform for multidimensional information By integrating the functional metabolic imaging capabilities of PET with the superior soft-tissue resolution and multiparametric functional imaging of MRI within a single device, PET/MRI enables multiscale and multidimensional characterization of tumor biological behavior (16). The MRI component not only provides detailed anatomical structures but also utilizes diffusion-weighted imaging (DWI) to quantify the Brownian motion of water molecules, reflecting cellular density; employs dynamic contrast-enhanced (DCE-MRI) to assess vascular permeability and perfusion; and leverages magnetic resonance spectroscopy (MRS) to noninvasively detect the concentrations of specific metabolites (17, 18). The PET component involves injecting radiolabeled tracers (such as 18F-FDG, 68Ga-PSMA, 18F-FET, etc.) to target key biological processes like glucose metabolism, specific receptor expression, or amino acid transport, thereby revealing the molecular phenotype of the tumor (19–21). The development of next-generation PET tracers continues to expand imaging capabilities; for example, Huang et al. recently reported 64Cu-NOTA-Ivonescimab, a novel ImmunoPET tracer targeting VEGF-A in colorectal carcinoma, demonstrating specific tumor uptake of 13.73 ± 0.95%ID/g at 48 hours post-injection (22). In clinical practice, PET/MRI has demonstrated superior value over positron emission tomography/computed tomography (PET/ CT) in the diagnosis, staging, and treatment response assessment of soft-tissue occupying lesions, including brain tumors, prostate cancer, gynecological malignancies, and hepatobiliary-pancreatic tumors (23–26). For instance, in glioblastoma, 18F-FET PET/MRI can effectively differentiate between tumor recurrence and radiation necrosis, with diagnostic accuracy significantly surpassing that of MRI alone (27). In prostate cancer, 68Ga-PSMA PET/MRI has emerged as the preferred method for localizing lesions following biochemical recurrence, profoundly transforming clinical decisionmaking pathways (28). However, the inherent limitations of PET/ MRI cannot be overlooked. While the MRI component provides high structural resolution, the overall system’s ability to characterize small lesions (<5 mm) is often constrained by the physical spatial resolution of the PET component. This limitation primarily affects metabolic sensitivity and quantification, potentially leading to the underestimation of activity in micrometastases due to partial volume effects (29). Additionally, tracers may exhibit non-specific uptake, leading to false-positive results. More importantly, PET/ MRI is inherently a local imaging technique, with its field of view confined to the scanned area. This makes it difficult to comprehensively assess systemic tumor burden and overall biological status, creating a blind spot in understanding the full evolutionary trajectory of the cancer. 2.2 Liquid biopsy: a molecular window for non-invasive dynamic monitoring Liquid biopsy refers to the non-invasive monitoring of tumors by analyzing tumor-derived materials in body fluids such as blood, urine, saliva, and cerebrospinal fluid (CSF). Its core components include ctDNA, CTCs, and exosomes (9, 30, 31) (Figure 1). ctDNA is fragmented DNA released into the bloodstream from apoptotic or necrotic tumor cells, carrying somatic mutations, copy number variations, and methylation profiles highly consistent with those of the primary tumor.

  Multimodal data fusion: integrating PET/MRI and liquid biopsy for a holistic view of cancer biology · 2026 · DOI

• Based on the aforementioned multidimensional complementary mechanisms, the multimodal data fusion of PET/MRI and liquid biopsy demonstrates significant value throughout the entire clinical diagnosis and treatment process (Figure 3). From early diagnosis and risk stratification to treatment efficacy evaluation, drug resistance monitoring, and even long-term follow-up and recurrence early warning, the two modalities can contribute with differential decision weights at different stages, collectively establishing a dynamic and precise tumor management system. 4.1 Early diagnosis and risk stratification In early cancer screening, liquid biopsy (such as multi-cancer detection based on ctDNA methylation signatures) can serve as efficient preliminary screening tools. However, a key challenge in general screening settings is that the low prevalence of cancer can limit the positive predictive value (PPV) of these tests, which, despite high specificity, leads to a risk of false-positive results (59). Whole-body PET/MRI technology provides crucial secondary verification and stratification capabilities. By integrating metabolic imaging (such as 18F-FDG PET) with high-resolution anatomical and functional MRI, this technology enables systemic evaluation of individuals with positive liquid biopsy results: accurately localizing suspicious lesions, distinguishing between benign and malignant conditions through multi-parameter analysis (e.g., SUVmax, ADC values), and offering anatomical guidance for subsequent interventions (74). This sequential “liquid biopsy-based initial screening followed by imaging-based precise screening” model has shown promise in early-phase and prospective cohort studies (e.g., the PATHFINDER study (59)) to improve the specificity and efficiency of screening. Recent large-scale prospective studies have further advanced this field: the K-DETEK study validated a multimodal ctDNA-based MCED test in 9, 057 asymptomatic individuals, demonstrating 70.8% sensitivity and 99.7% specificity (75).

  Multimodal data fusion: integrating PET/MRI and liquid biopsy for a holistic view of cancer biology · 2026 · DOI

• The transformation of breast cancer diagnostics from a single IHC-based assessment to a multi-omics, lon- gitudinal pipeline is the hallmark of modern precision Gholipour Maralan Egyptian Journal of Medical Human Genetics (2026) 27:32 medicine. While IHC remains essential for initial screen- ing and morphological context, it is no longer sufficient for the management of complex, evolving disease. Future directions: Interventional genomics: Clinical trials must now prioritize treating patients based on MRD status rather than waiting for radiological recurrence. Multi-modal AI integration: The future lies in combining "spatial transcriptomics" with serial liquid biopsy to create "digital twins" of patient tumors. Realizing this requires addressing significant technical challenges in data integration and demonstrating clinical utility. Equitable precision medicine: Standardizing bioinformatics and reducing sequencing costs is vital to bridge the "genomic divide" and ensure global access. Expanding the multi-omic lens: Future research must integrate epigenomics, metabolomics, and spatial transcriptomics to fully map the breast cancer landscape beyond genomics and transcriptomics.

  The multi-omic transformation of breast cancer diagnostics: a comprehensive narrative of the transition from immunohistochemistry to liquid biopsy and next-generation sequencing · 2026 · DOI

• These epigenetic modifications hint at pivotal roles in DMG pathogenesis, yet effective therapeutic strategies remain elusive, with median survival rates stagnant at approximately one year.

  DIPG-86. INVESTIGATING THE TUMOR ENVIRONMENT AND SIGNALING DEPENDENCIES DRIVING DIFFUSE MIDLINE GLIOMA TUMOR INITIATION, PROGRESSION, AND RESISTANCE: A COMPREHENSIVE APPROACH LEVERAGING BULK MULTIOMICS DECONVOLUTION AND SINGLE-CELL OMICS VALIDATION · 2024 · DOI

• While modern chimera prediction software allows for the fast and accurate identification of chimeric RNAs from RNA sequencing data, investigations separating therapeutically relevant transcripts from “transcriptional noise” remain lacking.

  Abstract 4351: Differential dependency mapping of chimeric RNAs across cancer reveals a new landscape of functional fusion transcripts · 2024 · DOI

• While IOBR 2.0 integrates transcriptomic and genomic data to explore the impact of the TME on patient phenotypes, it still has some limitations. First, the genomic analysis tools in IOBR 2.0 are relatively limited, only supporting basic analyses.

  Enhancing immuno-oncology investigations through multidimensional decoding of tumor microenvironment with IOBR 2.0 · 2024 · DOI

• The paper acknowledges significant challenges in protecting patient privacy within genomic medicine systems that integrate mobile health monitoring, blood biomarker testing, and genome information, but does not specify technical solutions, regulatory frameworks, or validation approaches for ensuring privacy-preserving integration of these personal genomic datasets.

  The era of genomic medicine · 2013 · DOI

• While the paper emphasizes that transcriptomic and proteomic analyses should enable progress in treating currently intractable neurological disorders like Alzheimer's disease, it does not specify what phenotyping improvements or biomarker combinations are required to distinguish disease subtypes or predict therapy responsiveness in neurodegenerative conditions.

  The era of genomic medicine · 2013 · DOI

• The paper proposes integrating human genomic information with monitoring of gut bacterial genomes and viral infections to enhance preventative medicine through personal health data, but provides no concrete framework for standardizing the analysis of these multi-omics datasets or establishing causative links between microbiome composition changes and disease risk in individual patients.

  The era of genomic medicine · 2013 · DOI

• The paper identifies uncertainty about the relative contributions of environment versus genetics in cancer etiology and acknowledges that for multiple western diseases (diabetes, arthritis, psychoses, dementia), the nature-nurture balance remains poorly defined. Establishing quantitative relationships between DNA methylation changes, chromatin modifications, and specific environmental exposures (diet, stress, toxins) in these disease contexts requires systematic epidemiological studies integrating genomic and environmental monitoring.

  The era of genomic medicine · 2013 · DOI

• PacBio system currently delivers 2 to 3 kb library inserts on average, although the polymerase can deliver longer reads, indicating a gap between technological capability and practical implementation.

  Bridging genomics technology and biology · 2013 · DOI

Most-cited papers in Cancer Genomics and Diagnostics

• Analysis of Circulating Tumor DNA to Monitor Metastatic Breast Cancer · New England Journal of Medicine · 2013 · 1,944 citations
• A Cell-free DNA Blood-Based Test for Colorectal Cancer Screening · New England Journal of Medicine · 2024 · 391 citations
• Recommendations for the use of next-generation sequencing (NGS) for patients with advanced cancer in 2024: a report from the ESMO Precision Medicine Working Group · Annals of Oncology · 2024 · 350 citations
• Mitochondrial Genome Instability and mtDNA Depletion in Human Cancers · Annals of the New York Academy of Sciences · 2005 · 221 citations
• ctDNA-based molecular residual disease and survival in resectable colorectal cancer · Nature Medicine · 2024 · 196 citations
• Insights for precision oncology from the integration of genomic and clinical data of 13,880 tumors from the 100,000 Genomes Cancer Programme · Nature Medicine · 2024 · 173 citations
• An atlas of epithelial cell states and plasticity in lung adenocarcinoma · Nature · 2024 · 158 citations
• Priming agents transiently reduce the clearance of cell-free DNA to improve liquid biopsies · Science · 2024 · 154 citations
• Origins and impact of extrachromosomal DNA · Nature · 2024 · 152 citations
• ClinVar: updates to support classifications of both germline and somatic variants · Nucleic Acids Research · 2024 · 140 citations

Most recent work

• Benchmarking long-read RNA-sequencing technologies with LongBench: a cross-platform reference dataset profiling cancer cell lines with bulk and single-cell approaches · bioRxiv · 2026
• Gene mutant dosage is associated with prognosis and metastatic tropism in 60,000 clinical cancer samples · medRxiv · 2026
• Blood‐based cell‐free DNA and multitarget stool RNA screening tests to detect colorectal cancer: A systematic review · Cancer · 2026
• Comprehensive profiling and clinical utility of CSF-derived cell-free DNA/RNA for evaluation of solid and hematologic malignancies affecting the central nervous system. · Journal of Clinical Oncology · 2026
• Reconstructing clone-resolved transcriptional programs from bulk tumor sequencing · bioRxiv · 2026
• Antisense oligonucleotide targeting TARDBP-EGFR splicing axis inhibits progression of oral squamous cell carcinoma through ABCA1-regulated cholesterol efflux · International Journal of Oral Science · 2026
• Methylation entropy as a novel dimension in liquid biopsy: enhanced multimodal framework for cancer detection and tissue-of-origin classification · Journal of Translational Medicine · 2026
• Age-related genomic characterization and therapeutic targets in Chinese breast cancer: insights from prospective targeted sequencing and clinical data analysis · BMC Medicine · 2026
• Critical Evaluation of Treatment Response, Driver Mutations, and Circulating Tumor DNA as Markers of Tumor Biology in Colorectal Liver Metastasis · Cancers · 2026
• The Circular DNA Size Code · Annals of Rejuvenation Science · 2026

Related topics in Biochemistry, Genetics and Molecular Biology