Seeing Disease Earlier, Faster, and More Clearly
A central pillar of this work is the recognition that health trajectories diverge long before diagnosis. By analyzing large-scale longitudinal clinical data, the team developed biological clocks that quantify how fast the body is developing or aging across the full life cycle, rather than relying on chronological age alone [1]. These models demonstrated strong predictive power for future disease incidence over long time horizons, enabling earlier risk stratification.
This concept is reinforced by multimodal imaging studies showing that biological age can be estimated with high precision—errors under two years—by integrating signals from different organs [2]. The idea extends to reproductive health: retinal imaging revealed a measurable “age gap” associated with ovarian reserve and menopause timing, linking ocular features to systemic reproductive aging in women [3]. In younger populations, predictive modeling of myopic progression identified children at risk of high myopia years before severe vision loss develops, enabling earlier intervention [4].
From anticipating risk, the research moves naturally toward detecting disease directly and efficiently. In pathology and radiology, annotation-free vision–language modeling enables localization of disease-relevant regions in medical images without manual labeling, allowing scalable analysis across organs and imaging modalities [5].
In the operating room, dynamic optical imaging combined with data-driven analysis distinguished malignant from non-malignant tissue in approximately three minutes, achieving diagnostic accuracy above 97% in prospective breast cancer cohorts—compared with roughly 30 minutes for conventional frozen-section workflows [6]. These advances focus on speed, clarity, and clinical usability, supporting decision-making when time is critical.
Detection alone is not enough; impact depends on understanding which signals matter for which patient. Several studies examined how the tumor microenvironment shapes outcomes. Multimodal fusion approaches uncovered previously hidden immunotyping heterogeneity in breast cancer, improving risk stratification beyond standard clinicopathologic categories [7].
In immunotherapy, multicohort analyses combined with single-cell sequencing showed that M1 macrophages and the CXCL9 signaling axis are strongly associated with response to immune checkpoint inhibitors across multiple cancer types, with predictive models outperforming conventional biomarkers such as PD-L1 [8]. In lung cancer, integrating imaging and clinical data enabled molecular phenotyping and prognostic prediction, linking non-invasive measurements to long-term outcomes [9].
Amplifying Weak Signals for Early Cancer Detection
Early detection represents another major impact pathway. Using blood-based cell-free DNA methylation patterns, the team demonstrated detection of early-stage ovarian cancer with around 80% sensitivity and 95% specificity in large validation cohorts, and translated key findings toward practical blood-test development [10]. Across these studies, learning-based models act as the connective tissue—integrating weak, sparse signals and revealing patterns too subtle for manual interpretation.
Underlying these diverse applications is a systems-level vision. Foundation models such as Meta-GP illustrate how electronic health records and multimodal imaging can be unified to address unmet clinical needs such as rare disease diagnosis and emergency care [11]. Complementary work on self-improving generative models shows how medical AI systems can refine performance over time and support downstream tasks through high-quality synthetic data [12]. Finally, a foundational architecture for AI agents outlines how these tools can be orchestrated into workflow-aware systems that plan, act, and learn safely within real clinical environments [13].
Conclusion: From Reaction to Anticipation
Taken together, Prof. Zhang Kang’s research presents a coherent trajectory: from early risk measurement, to rapid and precise disease detection, to biologically informed treatment decisions, and ultimately toward adaptive medical intelligence. The collective impact is a shift from reactive medicine to anticipatory, precision care, where insight arrives earlier—and clinical decisions improve as a result.
*Notes: This article provides research teasers for each reference to showcase the novelties
[1] K. Wang et al., “A full life cycle biological clock based on routine clinical data and its impact in health and diseases,” Nat. Med., 2025, doi: 10.1038/s41591-025-04006-w
[2] J. Wang et al., “Accurate estimation of biological age and its application in disease prediction using a multimodal image Transformer system,” Proc. Natl. Acad. Sci. U.S.A., vol. 121, no. 2, p. e2308812120, Jan. 2024, doi: 10.1073/pnas.2308812120
[3] H. Miao, S. Liu, Z. Wang, et al., “Artificial intelligence-derived retinal age gap as a marker for reproductive aging in women,” npj Digital Medicine, vol. 8, Art. no. 367, 2025, doi: 10.1038/s41746-025-01699-8.
[4] J. Li et al., “Accurate prediction of myopic progression and high myopia by machine learning,” Precision Clinical Medicine, vol. 7, Art. no. pbae005, Mar. 2024, doi: 10.1093/pcmedi/pbae005
[5] H. Yang et al., “A multimodal vision–language model for generalizable annotation-free pathology localization,” Nature Biomedical Engineering, vol. 9, p. 231, Mar. 2025, doi: 10.1038/s41551-025-01574-7
[6] S. Zhang et al., “Potential rapid intraoperative cancer diagnosis using dynamic full-field optical coherence tomography and deep learning: A prospective cohort study in breast cancer patients,” Science Bulletin, vol. 69, pp. 1748–1756, 2024, doi: 10.1016/j.scib.2024.03.061.
[7] Y. Yu, G. Cai, R. Lin, et al., “Multimodal data fusion AI model uncovers tumor microenvironment immunotyping heterogeneity and enhanced risk stratification of breast cancer,” MedComm, vol. 5, no. 12, Art. no. e70023, Dec. 2024, doi: 10.1002/mco2.70023
[8] Y. Yu et al., “Unraveling the role of M1 macrophage and CXCL9 in predicting immune checkpoint inhibitor efficacy through multicohort analysis and single-cell RNA sequencing,” MedComm, vol. 5, p. e471, 2024, doi: 10.1002/mco2.471
[9] Y. Lu, F. Liu, Y. Yu, et al., “AI-enabled molecular phenotyping and prognostic predictions in lung cancer through multimodal clinical information integration,” Cell Reports Medicine, vol. 6, Art. no. 102216, Jul. 15, 2025, doi: 10.1016/j.xcrm.2025.102216
[10] G. Li et al., “Transformer-based AI technology improves early ovarian cancer diagnosis using cfDNA methylation markers,” Cell Rep. Med., vol. 5, p. 101666, Aug. 2024, doi: 10.1016/j.xcrm.2024.101666
[11] F. Liu et al., “MetaGP: A generative foundation model integrating electronic health records and multimodal imaging for addressing unmet clinical needs,” Cell Reports Medicine, vol. 6, p. 102056, Apr. 2025, doi: 10.1016/j.xcrm.2025.102056
[12] J. Wang et al., “Self-improving generative foundation model for synthetic medical image generation and clinical applications,” Nat. Med., vol. 31, pp. 609–617, 2025, doi: 10.1038/s41591-024-03359-y
[13] F. Liu et al., “A foundational architecture for AI agents in healthcare,” Cell Reports Medicine, vol. 6, p. 102374, Oct. 2025, doi: 10.1016/j.xcrm.2025.102374.
From Natural Products to Drug Candidates: A Two-Way Research Model for Future Therapies in Cardiovascular, Arthritis, and Cancer Diseases
Over the past two decades, under the leadership of Professor Yi-Zhun Zhu, researchers at the Macau University of Science and Technology have been focusing on a distinctive bidirectional research model for small molecule drug development. This integrated strategy combines two complementary approaches: “from bioactive natural compounds to novel target identification” and “from novel therapeutic target to new small-molecule development.” Moving beyond conventional linear drug discovery processes, this framework facilitates bidirectional innovation, advancing both fundamental researches and translational applications.
Drug discovery can be conceptualized as an exploration on a vast island. Traditional approaches often follow a single linear route toward therapeutic breakthroughs, namely, the treasure. In contrast, the bidirectional model offers greater flexibility. Investigations may begin with a “coin” already in hand—such as Leonurine (SCM-198), S-Propargyl-Cysteine (SPRC), the natural compounds functioning as a valuable starting point. By studying that coin closely, researchers uncover clues about its origins, which in turn point them toward new biological targets and therapeutic directions, including NOX4, SMYD3, JMJD3, HDAC6 and so on. Alternatively, the journey begins the other way around, with a “map” in hand—a promising biological mechanism such as HDAC6/MyD88/NF-κB signaling pathway or STAT3-NAV2 axis in arthritis, serving as a guide for rational drug design. In both cases, the two pathways converge, leading to synergistic advances in drug discovery.
This strategy has yielded significant milestones, including the industrial translation of first-in-class therapeutics, and continues to drive progress across multiple therapeutic areas. In particular, the model has guided three major research trajectories: starting with leonurine, a natural product with traditional cardiovascular applications that has been advanced into modern drug development; extending into SPRC, a synthetically designed modulator of hydrogen sulfide; and progressing to epigenetic drug discovery through SMYD3 inhibition, a path with implications for vascular aging and oncology. Collectively, these directions illustrate how atural product-derived insights and target-driven design can converge within a single framework to shape the next era of medicine.
From Small molecule-to-Target: Leonurine
The journey along the “small molecule-to-target” pathway began with leonurine, a compound derived from Leonurus japonicus, a plant used in traditional medicine for cardiovascular health and gynecological diseases. What makes Leonurine significant is not only its herbal origins but also its development into a scientifically validated compound with cardioprotective effects, anti-inflammatory properties, pro-angiogenic effects following cardiac injury, and potential applications beyond cardiovascular disease which accounts for 17.9 million annual deaths globally [1] [2,3]. Beyond cardiovascular applications, it has demonstrated efficacy in mitigating endometriosis via modulation of estrogen-mediated immune dysfunction [4], as well as anti-atherosclerotic and hepatoprotective activities [5,6]. The team has also advanced Leonurine into novel formulations, the team developed a sustained-release leonurine formulation using PLGA microspheres encapsulating drug nanocrystals (Leo-nano@MP), which significantly improved lipid metabolism and reduced dosing frequency via subcutaneous administration in a hyperlipidemic rat model, demonstrating high drug loading, prolonged release, and good biocompatibility [7].
This pathway exemplifies how a single bioactive natural product can unravel novel mechanisms and open new therapeutic avenues. Leonurine thus serves as a prototype for transforming traditional herbal compounds into first-in-class therapeutics with validated mechanisms of action.
The translational development of Leonurine is supported by multiple patents from Professor Zhu’s group, covering pharmaceutical formulations and delivery technologies. For example, these include nanoliposome and microsphere systems that enhance bioavailability and enable controlled release, broadening the therapeutic window for cardiovascular and inflammatory diseases. Their patents also highlight leonurine’s role in modulating lipid metabolism, PPARγ pathways, and its use in treating liver disease and inflammation-related disorders. Collectively, these innovations elevate Leonurine from a traditional herbal compound into a modern, patent-protected therapeutic entity with substantial clinical potential.
From Small molecule-to-Target: another example, SPRC and CSE
If Leonurine represents the journey from monomer to target, S-Propargyl-Cysteine (SPRC) illustrates the complementary path—from target to molecule. Here, the therapeutic target is cystathionine Gamma Lyase (CSE), an enzyme that produces the signaling molecule hydrogen sulfide (H₂S) from L-cysteine. SPRC acts as a donor of endogenous H₂S, an endogenous gasotransmitter known for its profound effects on vascular health, inflammation, and longevity. Based on this mechanistic insight, SPRC was designed as a synthetic amino acid derivative that modulates H₂S levels.
Based on Professor Zhu’s studies, SPRC has demonstrated wide-ranging benefits: regulating immune responses, protecting neurons after stroke, and promoting angiogenesis [8,9]. In rheumatoid arthritis (RA), which affects over 18 million people globally [10], SPRC has been shown to rebalance immune cells by inhibiting the JAK/STAT pathway, reducing inflammation and slowing joint damage [11]. Beyond exploring the pharmacological efficacy of SPRC, the team have also engineered SPRC-based nanocarriers and hybrid molecules for precision therapies against RA,
This small molecule-to-target pathway highlights how identifying a novel biological mechanism—in this case, CSE as the target of SPRC—can inspire the rational design of first-in-class therapeutic candidates with translational potential.
Patents filed by Professor Zhu and his team surrounding SPRC and related signaling pathways demonstrate the translational depth of this pathway by engineering innovative drug delivery platforms. Microsphere and mesoporous silica formulations enhance the stability and sustained release of SPRC, improving its therapeutic precision in diseases such as rheumatoid arthritis and ischemic injury. Additionally, their patents on endogenous H₂S donors for RA treatment underscore the progression from mechanistic insight to tangible therapeutic inventions. These intellectual property advances ensure SPRC’s positioning not just as a promising molecule, but as a scalable and clinically adaptable treatment strategy.
Target-to-Molecule: SMYD3 and Epigenetic Regulation
Extending the target-driven approach further, the researchers turned their attention to the epigenetic regulator SMYD3, a histone methyltransferase linked to vascular aging, cellular remodeling, and cancer progression. Unlike Leonurine’s herbal origins, SMYD3 inhibition begins firmly at the target level, with the goal of designing molecules that disrupt pathogenic gene regulation.
Professor Zhu’s research team showed that, through high-throughput screening and structure-based drug design, small molecules such as ZYZ384 were identified to reduce cancer cell proliferation by lowering H3K4 trimethylation on oncogenic promoters. In models of hepatocellular carcinoma, SMYD3 inhibition suppressed tumor growth, offering a new precision epigenetic strategy against cancer [14]. Beyond oncology, targeting SMYD3 also addresses vascular senescence, the progressive aging of blood vessels that underlies hypertension, atherosclerosis, and stroke [15,16].
This work demonstrates how the target-to-molecule pathway can move beyond classical signaling molecules into the realm of epigenetics, opening entirely new frontiers for precision therapies in both cancer and age-related vascular disease.
The translational impact of SMYD3 research is evidenced by patents from Professor Zhu’s team covering novel SMYD3 inhibitors and their use in treating cancer and vascular disorders. These patents protect chemical entities capable of selectively modulating histone methylation, establishing a foundation for first-in-class epigenetic therapeutics in oncology and cardiovascular medicine.
The Bidirectional Model: Advancing Integrated Drug Discovery
Together, these approaches illustrate a cycle of discovery that moves beyond the conventional linear model of drug development. By leveraging natural products for target deconvolution and employing target-based rational design for new chemical entities, this bidirectional framework bridges traditional knowledge and contemporary science, integrating basic research with clinical translation. This model provides a robust and efficient strategy for developing small-molecule therapies that address major unmet medical needs in cardiovascular disease, arthritis, and cancer.
*Notes: This article provides research teasers for each reference to showcase the novelties
List of Patents
Preparation Method and Application of Leonurine Sustained-Release Microspheres
Preparation Method and Application of a Small-Molecule Inhibitor of Histone Methyltransferase SMYD3
Preparation Method of SCM198 Gel for Treating Skin Injuries
Microsphere Formulation Loaded with S-Propargyl-Cysteine and Its Preparation Method
Mesoporous Silica Formulation Loaded with S-Propargyl-Cysteine and Its Preparation Method
Endogenous Hydrogen Sulfide Sustained-Release Formulation: Preparation Method and Application
References
[1] https://www.who.int/health-topics/cardiovascular-diseases
[2] S. Luo, S. Xu, J. Liu, F. Ma, and Y. Z. Zhu, “Design and synthesis of novel SCM-198 analogs as cardioprotective agents: Structure-activity relationship studies and biological evaluations,” European Journal of Medicinal Chemistry, vol. 200, p. 112469, Aug. 2020, doi: 10.1016/j.ejmech.2020.112469.
[3] Z. Song, K. Song, Y. Xiao, H. Guo, Y. Zhu, and X. Wang, “Biologically Responsive Nanosystems Targeting Cardiovascular Diseases,” CDD, vol. 18, no. 7, pp. 892–913, Aug. 2021, doi: 10.2174/1567201818666210127093743.
[4] Y. Li et al., “Scm-198 alleviates endometriosis by suppressing estrogen-erα mediated differentiation and function of cd4+ cd25+ regulatory t cells,” Int. J. Biol. Sci., vol. 18, no. 5, pp. 1961–1973, 2022, doi: 10.7150/ijbs.68224.
[5] Y.-Y. Qiu, J. Zhang, F.-Y. Zeng, and Y. Z. Zhu, “Roles of the peroxisome proliferator-activated receptors (Ppars) in the pathogenesis of nonalcoholic fatty liver disease (Nafld),” Pharmacological Research, vol. 192, p. 106786, Jun. 2023, doi: 10.1016/j.phrs.2023.106786.
[6] M. Huang et al., “The multifaceted anti-atherosclerotic properties of herbal flavonoids: A comprehensive review,” Pharmacological Research, vol. 211, p. 107551, Jan. 2025, doi: 10.1016/j.phrs.2024.107551.
[7] Song Z, Meng S, Tang Z, Yang X, He Y, Zheng Y, Guo H, Du M, Zhu Y, Wang X. Injectable leonurine nanocrystal-loaded microspheres for long-term hyperlipidemia management. Biomater Sci. 2023 Jun 27;11(13):4713-4726. doi: 10.1039/d3bm00211j.
[8] S. Liu et al., “Endogenous hydrogen sulfide regulates histone demethylase JMJD3-mediated inflammatory response in LPS-stimulated macrophages and in a mouse model of LPS-induced septic shock,” Biochemical Pharmacology, vol. 149, pp. 153–162, Mar. 2018, doi: 10.1016/j.bcp.2017.10.010.
[9] Y. Xiong et al., “ZYZ-803, a novel hydrogen sulfide-nitric oxide conjugated donor, promotes angiogenesis via cross-talk between STAT3 and CaMKII,” Acta Pharmacol Sin, vol. 41, no. 2, pp. 218–228, Feb. 2020, doi: 10.1038/s41401-019-0255-3.
[10] https://www.who.int/news-room/fact-sheets/detail/rheumatoid-arthritis
[11] W. Cai et al., “S-propargyl-cysteine attenuates temporomandibular joint osteoarthritis by regulating macrophage polarization via Inhibition of JAK/STAT signaling,” Mol Med, vol. 31, no. 1, p. 128, Apr. 2025, doi: 10.1186/s10020-025-01186-6.
[12] Q. Ding et al., “Signaling pathways in rheumatoid arthritis: implications for targeted therapy,” Sig Transduct Target Ther, vol. 8, no. 1, p. 68, Feb. 2023, doi: 10.1038/s41392-023-01331-9.
[13] Z. Tang et al., “Neutrophil‐Mimetic, ROS Responsive, and Oxygen Generating Nanovesicles for Targeted Interventions of Refractory Rheumatoid Arthritis,” Small, vol. 20, no. 20, p. 2307379, May 2024, doi: 10.1002/smll.202307379.
[14] Q. Ding et al., “A novel small molecule ZYZ384 targeting SMYD3 for hepatocellular carcinoma via reducing H3K4 trimethylation of the Rac1 promoter,” MedComm, vol. 5, no. 10, p. e711, Oct. 2024, doi: 10.1002/mco2.711.
[15] Q. Ding, C. Shao, P. Rose, and Y. Z. Zhu, “Epigenetics and Vascular Senescence–Potential New Therapeutic Targets?,” Front. Pharmacol., vol. 11, p. 535395, Sep. 2020, doi: 10.3389/fphar.2020.535395.
[16] Z. Lin et al., “Discovery of deoxyandrographolide and its novel effect on vascular senescence by targeting HDAC1,” MedComm, vol. 4, no. 5, p. e338, Oct. 2023, doi: 10.1002/mco2.338.
