Parkinson’s disease diagnosis may soon undergo a significant shift, according to recent research published in Nature. A comprehensive multi-omics profiling study led by researchers at Beijing Hospital, including Dongdong Wu, Xinxin Ma, Huimin Chen, Huijing Liu, and Jing He, has identified neuroinflammation-related genes and exosomal microRNAs as robust diagnostic signatures for the neurodegenerative condition.
The study, posted on July 15, 2026, addresses a critical clinical gap: while neuroinflammation acts as a pivotal driver amplifying the pathogenic cascade within the Parkinsonian brain, the specific molecular markers connecting this inflammation to disease prognosis have remained unclear. By integrating multiple layers of transcriptomic data, the research team established a precise, machine-learning-validated framework to identify five core biomarkers capable of distinguishing Parkinson’s patients from healthy controls.
Multi-Omics Profiling Reveals 426 Differentially Expressed Genes
To uncover the molecular underpinnings of Parkinson’s disease, the Beijing Hospital researchers analyzed publicly available transcriptomic datasets from the Gene Expression Omnibus (GEO) repository, encompassing microarrays (GSE75249 and GSE22491), high-throughput RNA-seq (GSE269775), and single-cell RNA sequencing (GSE223138) profiles. According to the findings, bulk RNA-seq data analysis revealed 426 differentially expressed genes (DEGs) between Parkinson’s patients and healthy controls, which included 215 upregulated and 211 downregulated genes.
Volcano plot analysis highlighted significant upregulation of genes such as PTGDS, CD300E, LOC442245, PRKAG3, COL13A1, LRFN2, ZNF750, LOC728543, CAPN6, and FOLR3. Conversely, marked downregulation appeared in BTNL8, TRPM6, CACNG6, CLC, LOC654433, HBD, EPB42, SELENBP1, ALAS2, and CCDC27. When researchers cross-referenced these transcriptomic variations with known neuroinflammation-related genes, they identified 35 overlapping candidates—including PTGDS, TREM2, CCL4, AGTR1, and TACR1—serving as candidate mediators linking neuroinflammation and disease pathogenesis.
Machine Learning Framework Identifies Five Core Biomarkers
To screen for reliable diagnostic signatures, the research team deployed an integrative machine learning framework incorporating ten distinct algorithms and 101 combinations. Tested across the GSE75249 dataset via 10-fold cross-validation and subsequently validated in the GSE22491 dataset, the glmBoost combined with Ridge regression emerged as the optimal predictive algorithm.
This computational pipeline narrowed the candidate pool down to five core diagnostic biomarkers: PTGDS, RTN3, MAG, PROK2, and CNTNAP2. Receiver operating characteristic (ROC) analysis demonstrated that the area under the curve (AUC) values for all five biomarkers exceeded 0.7 in both the training and validation cohorts. These figures confirm their robust, independent diagnostic potential. To translate these genomic discoveries into clinical utility, the investigators constructed a quantitative diagnostic nomogram complete with confusion matrices and calibration curves to evaluate predicted versus observed risks.
Single-Cell Insights Reveal Monocyte Dominance and Intercellular Signaling
Beyond bulk tissue analysis, single-cell RNA sequencing data examined in the study (GSE223138) mapped out seven distinct cell clusters within the cellular landscape of Parkinson’s disease. Monocytes were identified as the predominant cell population. Pseudotime trajectory analysis tracked the developmental dynamics of this major monocyte lineage, while CellChat analysis exposed active intercellular signaling, specifically highlighting the monocyte ligand RETN.
Functional enrichment analyses via Gene Ontology (GO) annotation and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway mapping tied these neuroinflammation candidates to biological processes such as the regulation of neuroinflammatory responses and tumor necrosis factor production. Furthermore, enriched pathways included the calcium signaling pathway, neuroactive ligand-receptor interactions, and the PI3K-Akt signaling pathway, providing a comprehensive map of the molecular disruptions characterizing the Parkinsonian brain.