Major depressive disorder (MDD) exhibits such pronounced heterogeneity that conventional symptom-based diagnoses and monoamine-focused treatments have proven largely inadequate, pushing the field toward what we now call precision psychiatry. In this review, we map out stratification approaches rooted in mechanisms spanning neural circuits, immune function, metabolic regulation, and the gut-brain axis. We also discuss how multi-omics, neuroimaging, and digital phenotyping can be integrated to construct continuous, biological grounded disease subtypes. Building from this groundwork, precision drug development has increasingly turned its attention to targets with genetic validation, therapeutics tailored to specific biological profiles, and adaptive trial designs that leverage biomarker-guided patient selection. Making this vision operational demands more than scientific insight as it requires weaving predictive algorithms into closed-loop care systems and everyday clinical practice, all while keeping health equity front and center so that technological progress does not end up widening gaps in access and outcomes. Real progress in MDD care will only emerge when we manage to blend deep mechanistic understanding with intelligent data analysis and genuinely inclusive implementation.

Identifying the molecular targets of bioactive natural products (NPs) remains a critical bottleneck in drug discovery and the modernization of Traditional Chinese Medicine (TCM). Conventional target deconvolution strategies, such as affinity chromatography and activity-based protein profiling (ABPP), are often constrained by strict chemical modification requirements, incomplete proteome coverage, and difficulties in capturing weak or transient interactions. To overcome these limitations, this review highlights degradation-based protein profiling (DBPP), an emerging chemoproteomic strategy that applies proteolysis targeting chimera (PROTAC) technology to target identification. Unlike traditional occupancy-driven methods, DBPP employs an event-driven mechanism, converting complex physical binding events into amplified, detectable protein depletion signals via the ubiquitin-proteasome system. We systematically outline the core framework of DBPP, with particular emphasis on the rationally designed 'PROTAC toolbox', the probe-mixed strategy, and the dual-path orthogonal validation that integrates quantitative degradation proteomics with immunoprecipitation-mass spectrometry (IP-MS). Representative case studies involving NPs such as celastrol and artemisinin illustrate the potential of DBPP to identify elusive targets, including non-catalytic and weak-binding proteins, as well as reducing false positives through orthogonal validation. Finally, we discuss current methodological limitations and explore possible prospects for integrating DBPP with artificial intelligence (AI), single-cell omics, and spatial transcriptomics techniques as an effective way toward deciphering the polypharmacology of complex NPs.

Dual KRAS-epidermal growth factor receptor (EGFR) inhibition holds promise for KRAS-mutant colorectal cancer (CRC), yet drug resistance remains a key hurdle. This study by Zhang et al. identifies the SMAD family member 1 (SMAD1)-fibroblast growth factor receptor 3 (FGFR3) axis as the driver of Paneth-like lineage plasticity that mediates such resistance and demonstrates that FGFR3 targeting restores drug sensitivity and synergizes with dual pathway inhibition in preclinical models^[1].

How oligodendrocytes (OLs) perceive the geometric features of axons and convert them into precise myelin structures remains unclear. The study by Dereddi et al. showed that the ion channel transmembrane protein 63A (TMEM63A) mediates the conversion of membrane tension at the contact site through activating intracellular Ca²⁺ signaling. This work demonstrates that TMEM63A is crucial for tuning myelin geometry, including thickness, length, and compatibility with axons. The deficiency of TMEM63A aligns with the clinical phenotype of human hypomyelinating leukodystrophy-19 (HLD19) syndrome. These findings provide new potential therapeutic targets for the intervention of diseases related to hypomyelinating disorders.

Chimeric antigen receptor (CAR)-based adoptive cell therapies have displayed outstanding efficacy in hematological malignancies; in spite of this, success in solid tumors remains limited. Natural killer (NK) cells have arisen as an encouraging alternative to T cells for cancer immunotherapy owing to their innate cytotoxic capability, negligible risk of graft-vs-host disease (GVHD), and lowered cytokine release syndrome (CRS). Regardless of these improvements, NK cell therapies face substantial challenges in the solid tumor setting, with an immunosuppressive tumor microenvironment (TME), inadequate persistence, and poor metabolic fitness. In a recently published study, Yang et al. utilized a CRISPR-based Synergistic Activation Mediator (SAM) screen to find novel genetic targets capable of improving NK cell anti-tumor function. Using an in vivo screening approach with over 70,000 guide RNAs in an NK92 cell line model bearing colorectal tumors, the investigators identified olfactory receptor OR7A10 as a top-ranked candidate gene. Consequent validation in primary NK cells and third-generation CAR-NK constructs demonstrated that OR7A10 overexpression significantly boosted cytolytic activity among multiple solid tumor types, enriched metabolic fitness through increased oxidative phosphorylation and mitochondrial biogenesis, and conferred resistance to the hostile TME. These results acknowledged OR7A10 as a novel molecular target for engineering next-generation CAR-NK cell products with improved anti-tumor efficacy in solid malignancies[1].

Metabolites are involved in almost all fundamental biological processes. The identification of their protein targets is crucial for elucidating non-canonical signaling roles and evaluating their therapeutic potential. In recent years, due to the continuous development of chemical proteomics, atypical phenotypic functions of classic metabolites have been continuously discovered. However, the discovery of endogenous metabolites for their disease-related functions is still progressing slowly. To accelerate the identification of disease-related targets for endogenous metabolites, we propose a new hypothesis: endogenous metabolites and their molecular targets are expected to have similar disease associations. Following this hypothesis, here we report the development of a novel deep-learning model called DeepETD, which integrates bioinformatics data and introduces an attention mechanism to predict functional targets of specific metabolite phenotypes. Using this model, we constructed a publicly accessible database named EMTDD containing potential targets for 3,382 common human endogenous metabolites. Overall, this study presents a new computational method and resource for endogenous metabolite target discovery as an important supplement to experimental methods such as chemical proteomics.

Many disease-associated proteins are difficult to target using conventional small molecules or biologics. RNA offers a promising therapeutic strategy because its structural motifs can create recognition sites for regulatory proteins and small molecules, enabling intervention upstream of proteins that are otherwise considered "undruggable". Recent advances in RNA-targeting therapeutics, particularly RNA-binding small molecules, have demonstrated the feasibility of selectively modulating RNA maturation, splicing, translation, and degradation. However, their broader clinical application is still limited by non-specific toxicity, off-target effects, stability concerns, and delivery challenges. Targeted protein degradation is a remarkable therapeutic modality that has been widely used to target undruggable pathogenic proteins. In relation to this, a currently emerging research field is that of ribonuclease-targeting chimeras (RiboTACs). Unlike the "occupancy-based" mode of small-molecule inhibitors (SMIs), RiboTACs recruit endogenous ribonucleases to selectively degrade disease-associated RNA transcripts through an event-driven mechanism, which enables efficient removal of pathogenic RNAs while offering advantages with regard to medicinal chemistry optimization compared with oligonucleotide-based approaches. In this perspective, we systematically summarize the classification of targetable RNAs, recent advances in RNA-targeting small molecules, and the design principles, applications, and future challenges of RiboTACs.

Biomedical knowledge graphs (KGs) provide a structured and traceable framework for target-oriented, AI-assisted drug discovery. By linking genes, proteins, diseases, phenotypes, compounds, pathways, assays, structures, adverse events, and clinical evidence, biomedical KGs can partially mitigate evidence fragmentation and support hypothesis generation across molecular, cellular, and clinical levels. This review summarizes how biomedical KGs are constructed, reasoned over, evaluated, and translated into practical drug-discovery tasks. We first outline major data resources, entity and relation extraction, knowledge fusion, quality control, and large language model-assisted KG construction, with emphasis on ontology grounding, evidence tracing and hallucination control. We then compare key KG reasoning paradigms, including symbolic rule-based reasoning, representation learning, path- and subgraph-aware reasoning, graph Transformer models, and KG-LLM hybrid reasoning. Their strengths and limitations are discussed in relation to interpretability, scalability, leakage risk, benchmark bias, and prospective validation. We further highlight the need to incorporate three-dimensional structural evidence, including protein structures, binding pockets, variants, conformational states, and ligand compatibility, so that graph-based relevance can be connected with chemical tractability. Finally, we discuss applications of biomedical KGs in target prioritization, druggability assessment, mechanism interpretation, drug-target interaction prediction, molecular generation, lead optimization, ADMET and safety assessment, drug repurposing, combination therapy, biomarker discovery, precision medicine, and drug-resistance modeling. Current limitations include heterogeneous data quality, incomplete causal evidence, static graph representations, privacy constraints, and insufficient experimental or clinical validation. Overall, biomedical KGs should be viewed as evidence-integration and prioritization engines rather than automatic proof-generating systems for drug discovery, guiding rigorous, iterative, and experimentally grounded translational decisions.

Polysaccharides, together with proteins and nucleic acids, are typically considered the three fundamental macromolecules essential for life. Unlike well-studied proteins and nucleic acids, polysaccharides remain poorly characterized. Their inherent structural heterogeneity makes them particularly challenging to study with conventional techniques. Artificial intelligence (AI) has emerged as a transformative technology in driving the paradigm shift of polysaccharide research to data-driven intelligence, thereby enabling efficient analysis of extensive data. Herein, we systematically review AI applications in polysaccharide research, mainly focusing on various stages in polysaccharide drug development. The limitations and outlooks are discussed as well, following the review of the advantages of AI in this field.

Site-specific degradation by targeting chimeras (TACs) has become a desirable alternative to occupancy-driven inhibition for central nervous system (CNS) diseases due to the possibility of completely deactivating pathogenic proteins, rather than complete blocking. The catalytic mechanism of TACs is dynamic, meaning that they can provide sustained therapeutic effects at low doses in clearing key pathogenic proteins associated with CNS diseases, while resisting the development of resistance mechanisms. However, the physicochemical properties of TACs pose a high delivery question for the brain, limiting brain bioavailability and impairing clinical progress. To address this issue, this review presents the structure-function relations of TAC modules, focusing on their therapeutic potential in various CNS diseases such as glioblastoma, Alzheimer's disease, and Parkinson's disease, as well as the role of novel nanocarriers and delivery platforms in overcoming biological barriers. The review is intended to inform future brain-penetrant TAC therapeutics.

Traditional Chinese Medicine (TCM), with a history of thousands of years, has been increasingly supported by modern clinical studies for its effectiveness in disease prevention, treatment, and rehabilitation. However, the complex, multi-component, and multi-target characteristics of TCM make it challenging to identify bioactive compounds and understand the molecular mechanisms underlying their effects. These features also complicate the standardization of clinical applications. However, investigation of TCM targets remains limited. Systems biology and network pharmacology have become valuable approaches for analyzing the complex multi-component and multi-target interactions of TCM. Omics-based approaches, including transcriptomics and single-cell sequencing, have further improved the ability to characterize gene expression changes and cell heterogeneity. Recently, artificial intelligence (AI) has emerged as a powerful tool for identifying potential targets within TCM. The continuous improvement of professional TCM databases supports the integration of information, including prescriptions, bioactive constituents, target interactions, disease associations, and pharmacological annotations, gradually building a more comprehensive data system for TCM research. These databases serve as key resources, allowing AI-based approaches to identify targets in TCM. This review summarizes recent progress in AI-assisted TCM target identification, discusses experimental strategies for validating computational predictions, and outlines the main methodological challenges and future directions.

Pulmonary arterial hypertension (PAH) is a fatal, progressive cardiovascular disease. Treprostinil (TRE) is a first-line drug for PAH treatment. However, the use of TRE for severe PAH is often limited by the risks and complexities of pain at the subcutaneous (SC) infusion site or fatal catheter-associated intravenous (IV) infections with external or implanted pump systems. We engineered an injectable in situ-forming TRE depot at physiological temperature to prolong vasodilation by loading drug microparticles into a methylcellulose-based gel. A single intramuscular (IM) injection significantly prolonged the drug's residence at the injection site, increased the half-life of elimination, and lowered the maximal plasma concentration. One injection induced pulmonary vasodilation lasting 3–4 d, improved hemodynamics, and reduced pulmonary vascular remodeling. A safety evaluation showed favorable biocompatibility and local tolerability, with a mild local inflammatory response resolving within 1 week. In conclusion, the injectable TRE depot is a promising alternative to pump delivery, reducing the administration burden and improving compliance and quality of life.