Design of Single-Atom Nanozymes for Precision Treatment of Erectile Dysfunction with Integrated Single-Cell RNA Sequencing and Machine Learning

An ML-guided iron nanozyme reduced oxidative stress in diabetic rat corpus cavernosum, restoring erectile function in an animal model

Journal: Advanced Science | Published: 2026-04-14 | Type: Journal Article | PMID: 41980199 Authors: Zhou Xiang, Zhang Xi, Chen Sihan et al. (Department of Urology, First Affiliated Hospital of Nanjing Medical University) Funding/COI: Wu Jieping Medical Foundation, China Postdoctoral Science Foundation, National Natural Science Foundation of China | COI: not disclosed

Summary

Diabetes-related ED responds poorly to first-line PDE5 inhibitors, and the oxidative stress mechanism is a plausible reason why. This paper uses single-cell RNA sequencing on human corpus cavernosum tissue to map that dysfunction, feeds the findings into machine learning models to screen a nanozyme database, and synthesizes a specific iron-based nanoparticle (Fe-DMOF) that mimics three antioxidant enzymes simultaneously. The nanoparticle was then tested in diabetic rodents, where it reduced penile oxidative stress and reversed downstream inflammatory changes.

Claims

• scRNA-seq on DMED patient corpus cavernosum tissue showed elevated ROS and downregulated GPx, CAT, and SOD expression compared to controls
• Two ML models trained on a curated nanozyme database predicted Fe-based single-atom nanozymes as the best match for this specific enzyme-deficiency profile
• Fe-DMOF simultaneously exhibited GPx-, CAT-, and SOD-mimicking enzymatic activity in vitro
• Treatment reduced ROS-induced histone lactylation modifications in corpus cavernosum cells
• Fe-DMOF reversed inflammatory differentiation of fibroblasts and macrophages in diabetic corpus cavernosum (cell and animal data)
• Erectile function improved in the diabetic animal model following Fe-DMOF administration (specific ICP/MAP ratios not reported in abstract)

Study Quality

This is a multi-stage preclinical study, not a clinical trial. The human data component is scRNA-seq on tissue samples from DMED patients — a valid discovery tool for identifying molecular targets, but the abstract does not report how many patients donated tissue or what their clinical characteristics were. The ML framework is methodologically novel: building a structured nanozyme database and training classifiers to predict enzyme-mimicking activity is a reasonable data-driven approach, though the database size and model performance metrics are not disclosed in the abstract. Synthesis and in vivo testing follow a logical pipeline, but all functional outcomes are in rodents.

The MeSH terms confirm both animal and human components, but the primary efficacy evidence is animal-derived. No head-to-head comparison with existing treatments (e.g., sildenafil in diabetic rats) is mentioned. The mechanism proposed — ROS-induced histone lactylation promoting inflammatory fibroblast/macrophage differentiation — is biologically plausible and recently described, which adds novelty but also means it is less independently validated.

Red Flags

• Sample size for human scRNA-seq tissue not disclosed in abstract — critical for assessing whether the transcriptomic findings are robust or driven by one or two outlier patients
• All efficacy data appear to be from a diabetic rodent model; human translation is entirely speculative at this stage
• No comparison condition against sildenafil or tadalafil in the animal model, making it impossible to contextualize any improvement
• COI not disclosed despite institutional industry-adjacent funding streams; absence of declaration is not the same as absence of conflict
• Fe-DMOF is a synthesized nanoparticle with no safety data reported — biodistribution, clearance, and off-target enzyme activity are unanswered questions
• Journal (Advanced Science, Wiley) is high-IF but known for publishing heavily engineered nanomaterial papers that rarely survive clinical translation

Strengths

• Uses actual human corpus cavernosum tissue rather than relying solely on animal models for the discovery phase
• scRNA-seq at single-cell resolution provides more granular mechanistic insight than bulk RNA approaches
• The ML-guided design framework is genuinely novel and potentially generalizable beyond this indication
• Identifies a specific downstream mechanism (histone lactylation as a consequence of ROS accumulation) that could serve as a biomarker or secondary target
• Addresses a real unmet clinical need: DMED patients have substantially worse outcomes on current first-line therapies

Verdict

Technically impressive but several steps removed from anything clinically actionable. The scRNA-seq + ML pipeline is the actual contribution here — it's a legitimate methodological advance in nanomedicine design, and the iron nanozyme it produced shows meaningful activity in a diabetic rodent model. But the human tissue dataset has undisclosed sample sizes, all functional outcomes are in rats, and Fe-DMOF has no safety profile. Read this as a proof-of-concept for a data-driven nanozyme discovery approach, not as evidence that iron nanoparticles treat erectile dysfunction. If the authors run a well-powered follow-up with head-to-head animal comparisons, proper toxicology, and a disclosed COI section, it becomes more interesting.