Proteomics enables systematic, context-dependent characterization of the proteome, and has emerged as a cornerstone of precision cancer medicine. Although the systematical analysis of molecular profiling has transformed oncology over the past two decades, substantial heterogeneity in therapeutic responses still persists among patients with similar genetic alterations, highlighting the limitations of static genomic information. By directly interrogating signaling pathways and regulatory networks of proteins and post-translational modifications (PTMs) that drive tumor initiation, progression, and therapy resistance, proteomics bridges the gap between genomic alterations and phenotypic outcomes. Recent advances in mass spectrometry have enabled low-cost, high-throughput, and high-resolution proteomics from bulk to single cells, providing unprecedented insights into tumor heterogeneity. Integrative analysis of multi-omics, including genomics, transcriptomics and proteomics data, facilitates the construction of multidimensional molecular landscapes that reveal novel biomarkers and therapeutic targets. In this review, we summarize recent advances in proteomics-based biomarker discovery, highlight emerging single-cell and spatial proteomics technologies, and discuss future directions for integrating multi-omics, clinical information, and artificial intelligence to accelerate clinical translation.

Colorectal cancer (CRC) ranks among the leading malignancies globally in both incidence and mortality. Treatment failure and disease recurrence are largely attributable to significant heterogeneity and acquired resistance to current therapies. Recent studies have extensively demonstrated that the initiation, progression, and recurrence of CRC are closely associated with the malignancy of intestinal stem cells (ISCs) and their derived cancer stem cells (CSCs). CSCs possess the characteristics of sustained self-renewal and multi-potent differentiation, constituting a key cellular population that sustains tumor growth, metastasis and drug resistance. Concurrently, CSCs evade the host immune system by reducing tumor antigen presentation, secreting immunosuppressive factors, and remodeling tumor microenvironment, thereby significantly limiting the clinical efficacy of immunotherapy. Consequently, immunotherapeutic strategies targeting ISCs and CSCs have emerged as a pivotal research direction for precision treatment of CRC. This paper systematically reviews recent advances in this field, discussing vaccine strategies based on CSC-specific antigens, bispecific antibodies (BsAbs) and antibody-drug conjugates (ADCs), CAR-T cells, and multimodal therapeutic approaches. Further, this paper summarizes the application of multi-omics technologies, spatial biology, and organoid models in elucidating the plasticity and drug resistance mechanisms of CSCs. We also discuss the potential role of gut microbial regulation in enhancing immunotherapy response. In summary, comprehensive immunotherapy strategies targeting ISCs and their ecological niches hold promise for overcoming current treatment bottlenecks in CRC, providing new theoretical foundations and practical pathways towards achieving long-term disease control.

Cancer prevention is, in practice, a prediction problem. Decisions about risk reduction, screening, and survivorship depend on forecasts of incident cancer, progression, recurrence, and treatment-related harms, and on whether earlier or more intensive action improves the benefit-harm balance. For prevention-oriented use, prediction tools should provide calibrated absolute-risk estimates over a defined follow-up period, be anchored to local incidence and competing mortality when relevant, and be embedded in explicit clinical or public health pathways in which prespecified thresholds trigger actionable steps. These steps may include intensified lifestyle support, eligibility assessment for chemoprevention, risk-adapted screening decisions regarding start age, interval, or modality, and referral for confirmatory evaluation. Using breast cancer as the most mature archetype, we illustrate how individualized risk estimates can be linked to multiple prevention levers while making the trade-offs of threshold-based strategies explicit, including induced follow-up procedures and the potential for overdiagnosis. We then synthesize requirements for prevention-ready prediction, spanning longitudinal follow-up with registry-linked outcome ascertainment, absolute-risk estimation anchored to local incidence and competing mortality, evaluation of calibration overall and in relevant subgroups, transparent accounting of threshold-induced follow-up procedures and harms, and planned recalibration and updating as baseline risk and practice patterns change. As cohort-enabled illustrations in the Health Examinees (HEXA) infrastructure, we describe a breast cancer prediction benchmark, CancerFree, a questionnaire-first, registry-anchored multi-cancer framework for site-specific risk-to-action specification, alongside an AI-derived diet layer that converts food frequency questionnaire data into interpretable dietary-pattern features for scalable prevention-oriented profiling. Across emerging data layers, added model complexity should be justified not by discrimination alone, but by demonstrable gains in calibrated absolute-risk estimation and in the benefit–harm balance of the pathways they are intended to inform.