In recent years, China's new round of institution reform has further optimized the drug regulatory system. Relevant departments and institutions involved in traditional Chinese medicine (TCM) regulation have been strengthened. TCM regulatory science, as an emerging interdisciplinary field, has received high regard and experienced rapid development, significantly enhancing TCM regulatory capabilities. Simultaneously, accelerated progress in emerging technologies and production innovation for TCM drug discovery, coupled with the implementation of the National Major Scientific and Technological Special Project for "Significant New Drugs Development" and its translational achievements, have led to a historic turning point in the development of innovative natural TCM drugs over the past five years. Driven by the dual engines of "regulatory science" and "policy restructuring", the development of new TCM drugs has entered a fast lane. Both the quantity and quality of investigational new drug (IND) and new drug application (NDA) registrations and approvals for new natural TCM drugs have shown rapid growth, effectively meeting the public's health demands for TCM products and unmet clinical needs of patients. This study focuses on the development of new TCM drugs during the significant historical phase from 2021 to 2025. It provides a comprehensive overview of new TCM and natural drug applications and regulatory reviews over the past five years, delves into the implementation of the National Drug Regulatory Science Action Plan, and highlights the importance of TCM regulatory science as an emerging interdisciplinary field in accelerating the creation of new TCM drugs. It systematically summarizes the effects of regulatory policies and regulations, the reform of TCM registration classification, specialized TCM registration provisions, and incentive measures such as the National Major Scientific and Technological Special Project for "Significant New Drugs Development". Based on an international perspective, it provides a focused review of recent highlights in TCM new drug development and regulation. This holds significant importance for promoting breakthroughs in TCM new drugs across more disease areas and advancing the international coordination of TCM regulation. The challenge faced in managing the registration of new TCM drugs lies in resolving the conflict between TCM theory and modern drug attributes, while balancing the rapid advancement of traditional medical theory and emerging technologies with the robustness of the drug regulatory framework. In the future, actively advancing research and translation in TCM regulatory science, innovatively establishing benefit-risk assessment systems and standards for new TCM drugs, and accelerating the development of a globally leading regulatory system with Chinese characteristics that aligns with the unique nature of TCM will be particularly crucial for global coordination of TCM regulatory policies, and the modernization and internationalization of TCM.

Enzymatic reactions play an emerging role in a broad spectrum of scientific and industrial applications. The inherent complexity of enzymes, such as their substrate specificity, conformational flexibility, and the vast diversity of reactions involved, poses substantial challenges for the advanced computational prediction of enzymatic reactions with desirable accuracy. Moreover, existing approaches are mostly tailored for a specific sub-task, such as substrate prediction or binding site annotation, which limits their applicability. In this study, we introduce ERAM, a task-agnostic multimodal learning framework capable of addressing a broad range of downstream applications with both accuracy and efficiency. ERAM aligns pre-trained molecular representations from Protein Language Model with the knowledge of enzyme catalysis by modeling enzymatic reactions as multi-relational data. In enzyme retrieval tasks, ERAM achieves an improvement of 28.31% in mean average precision compared with the state-of-the-art (SOTA) method, CREEP. In substrate prediction tasks, ERAM outperforms the SOTA method ESP, achieving average improvements of 35.53% and 22.97% in Matthews correlation coefficient across two datasets. Additionally, ERAM exhibits commendable interpretability by assigning higher attention weights to binding sites, resulting in lower false-positive rates (42.36%) and higher overlap scores (70.59%) in the unsupervised binding site prediction task compared to RXNAA Mapper. By learning embeddings of substrates, enzymes, and products within a unified knowledge graph latent space, ERAM demonstrates its potential as a versatile and effective tool for enzyme catalysis research.

Triple-negative breast cancer (TNBC) is a highly aggressive and heterogeneous subtype of breast cancer characterized by early metastasis, poor prognosis, and high recurrence rates. Targeting dysregulated PI3K/Akt/mTOR signaling and triggering anti-tumor immunity represent promising strategies for TNBC therapy. In this study, we report the discovery of a series of novel chromone derivatives as potent mTOR inhibitors by artificial intelligence-assisted drug design and structure-based drug design. The optimal compound, MT-44, was a highly selective mTOR inhibitor and showed no obvious binding activity to a broad panel of 200 kinases, and it exhibited nanomolar-level mTOR inhibitory and anti-TNBC cells proliferative activities. MT-44 effectively blocked the PI3K/Akt/mTOR signaling pathway and exerted robust anti-tumor efficacy in an MDA-MB-231 xenograft mouse model. Furthermore, MT-44 activated pattern recognition receptor TLR2 and upregulated the cGAS/STING signaling pathway, and reshaped the tumor microenvironment, thereby enhancing the tumor immune landscape. Collectively, our findings highlighted MT-44 as a highly selective and potent mTOR inhibitor with dual-targeted therapeutic and immunomodulatory effects, offering an appealing strategy for TNBC.

Biologic drugs, primarily comprising proteins and nucleic acids, have emerged as powerful therapeutic modalities; however, their discovery and optimization are often hindered by their inherent complexity. The advent of artificial intelligence (AI), particularly deep learning, is catalyzing a paradigm shift in this field, transitioning it from a process reliant on serendipity and laborious experimentation to a data-driven engineering discipline. This review systematically charts the co-evolution of AI methodologies and their transformative applications across the modern biologic drug development pipeline. We first outline AI’s methodological progression, from language models deciphering biological sequence grammar to structure prediction models like AlphaFold making macromolecular folds computationally accessible, and finally to generative models enabling de novo molecular creation. We then explore the practical impact of these technologies in two core phases: the de novo design of novel biologics with bespoke functions and the subsequent multi-parameter engineering and optimization of these candidates for clinical viability. While the potential is immense, significant strategic challenges remain, including the need to build a new AI-native experimental ecosystem and bridge the profound complexity gap between molecular-level predictions and systemic in vivo outcomes. Overcoming these obstacles will usher in a new era of AI-driven, automated closed-loop drug discovery.

Molecular glue degraders (MGDs) have emerged as a transformative modality in the field of targeted protein degradation (TPD), enabling the selective elimination of disease-relevant proteins, including those traditionally considered undruggable. Unlike bifunctional proteolysis-targeting chimeras (PROTACs), MGDs operate through monovalent architectures that induce protein-protein interactions (PPIs) between E3 ligases and neosubstrates, offering advantages in chemical simplicity, cell permeability, and target scope. However, MGD discovery remains serendipitously, and a translational framework that links rational design to predictable selectivity and tissue exposure is still lacking. In this review, we present an integrated framework for advancing next-generation MGDs through three critical dimensions: rational design, specificity optimization, and delivery systems. First, we examined cutting-edge strategies in MGD design, including covalent handle-based reprogramming, PPI-driven stabilization, and multi-site, multi-functional constructs. Second, we explored structure-guided engineering and chemoinformatic models, such as cereblon degron motifs, zone-based design and multiparameter optimization, to improve neosubstrate selectivity while minimizing off-target liabilities. Third, we summarized delivery platforms, including antibody-drug conjugates, nanoparticle-enabled systems, and folate-mediated targeting, which are primarily intended to improve tissue selectivity and targeted distribution, thereby promoting local tissue accumulation. Finally, we discussed emerging opportunities at the intersection of artificial intelligence, structural biology, and systems pharmacology for accelerating MGD discovery and clinical translation. Collectively, these interdisciplinary insights underscore the therapeutic promise of MGDs and lay the groundwork for their next-generation evolution in precision medicine.

Accurate prediction of protein-ligand complexes and binding affinity is critical for hit identification and optimization for structure-based drug design. Traditional docking simulates binding processes with searching algorithms guided by energy-scoring functions, which are quite computationally expensive and time-intensive. In contrast, deep learning approaches offer a cost-effective alternative, yet often generate conformations with limited physicochemical validity and fail to account for protein flexibility. To address these pitfalls, we propose FlowDock, a multitask framework enhanced by Bayesian Flow Networks. FlowDock simultaneously generates accurate protein-ligand complex structures and predicts binding affinity while incorporating protein conformational flexibility. By leveraging multimodal intramolecular representations with a deep equivariant generative model, our method iteratively refines complex in latent space, ensuring rapid and stable generation. Benchmark evaluations demonstrate that FlowDock achieves state-of-the-art performance in binding pose prediction, especially physical plausibility, and virtual screening capability, alongside reliable binding affinity predictions. By providing deeper molecular insights into dynamic protein-ligand interactions, FlowDock represents a robust tool for accelerating the rational development of therapeutics.

PI3Kγ represents a promising therapeutic target for its pivotal role in macrophage recruitment and polarization and its significant association with tumor invasion and metastasis. In this study, a series of novel indole-based PI3Kγ selective inhibitors were generated by machine learning combined with molecular hybridization. Intriguingly, the representative IHA-5f displayed picomolar-level potency and highly selective inhibition to PI3Kγ relative to PI3Kα/β/δ. Moreover, IHA-5f manifested prominent anti-melanoma activity in vitro and in vivo with no detectable visceral toxicity. Mechanistically, IHA-5f efficiently suppressed tumor cell proliferation and migration, and induced apoptosis by suppressing the PI3Kγ/AKT/NF-κB signaling axis. Concurrently, it restrained the M2 polarization of tumor-associated macrophages, thereby augmenting the antitumor immune response. This study underscores the potential for PI3Kγ inhibitors as immunomodulators and direct antitumor agents.

Despite extensive structural and functional characterization of the α7 nicotinic acetylcholine receptor, valuable structural insights into its interactions with conopeptides remain limited, thereby hindering the rational development of peptide-based modulators for this clinically important receptor subtype. Here, we present an integrated pipeline combining deep learning, structural biology, computational modeling and electrophysiology to accelerate the discovery and optimization of α7 nAChR-targeting conopeptides. To overcome data scarcity, we developed a deep learning model using the ESM-2 protein language framework, enabling efficient screening of 689 disulfide-poor conopeptides. This approach identified SS1, a novel antagonist of α7 nAChR, which was systematically optimized via structure-activity relationship studies to yield [△QP,S8R]SS1—a minimalist peptide with nanomolar potency (IC₅₀ = 49.2 nmol/L), enhanced selectivity, and improved stability. Cryo-EM and computational modeling resolved the 3.3 Å resolution structure of α7 nAChR bound to [S8R]SS1, revealing a unique binding mode stabilized by hydrogen bonds, hydrophobic interactions, and glycan contacts, while hybrid receptor conformations (closed/desensitized) elucidated its inhibitory mechanism. This work establishes a transformative deep learning-to-experiment framework for accelerating the discovery and optimization of nature-inspired peptide therapeutics.

Nanodrug delivery systems (NDDS) have demonstrated outstanding performance in drug delivery due to their efficient delivery capacity, targeting ability, and biocompatibility. However, the development of nanomedicines still heavily relies on the expertise of formulation scientists and extensive trial-and-error experiments. Despite the abundance of data in nanoscience, traditional biological research often struggles to effectively process, analyze, and utilize these datasets, limiting nanomedicine studies to a “one-to-one” approach. Against this backdrop, the rapid growth of artificial intelligence (AI) and machine learning (ML) offers a new paradigm for nanomedicine research. Unlike traditional statistical analyses and mathematical models, AI and ML provide deeper insights into big data, enhancing the efficiency of nanomedicine development while steering the field toward more intelligent and more precise research approaches. This review focuses on milestone studies that use ML to reshape nanomedicine research from a pharmaceutics perspective, highlighting how data-driven ML models can guide new directions in nanomedicine development.