While single-omics analyses of Parkinson's Disease (PD) have demonstrated their ability in revealing the underlying molecular mechanisms, they often fail to provide a comprehensive view of the complete disease mechanisms. In this study, we leveraged multi-omics data from 64 heterogeneous, well-phenotyped PD patients, generated plasma metabolomics data and Olink proteomics data together with the gut and saliva metagenomics data, and investigated the altered molecular mechanisms and their interactions in association with the severity of motor function disorders in PD patients. Based on our multi-omics approach, we identified a panel of 58 biomarkers comprising one clinical variable, 10 proteins, and 17 metabolites from plasma, 26 gut species, and 4 saliva species for PD severity. These biomarkers exhibited superior predictive performance for assessing PD severity compared to those derived from single-omics datasets. The predictive power of our machine learning models based on these biomarkers was validated using additional multi-omics data from the same group of PD patients after a 3-month follow-up. The contribution of each omics dataset was evaluated by both supervised and unsupervised machine learning approaches, highlighting the importance of plasma metabolomics in disease stratification. Our study unveiled disease-related molecular alterations across multiple omics datasets, offering potential diagnostic and therapeutic insights for PD. Moreover, it underpinned the significance of employing multi-omics analyses when studying complex diseases like PD.

Achieving minimally invasive and rapid detection is a crucial goal in modern medicine. The comprehensive characterization of the blood proteome holds great promise in advancing our understanding of disease etiology, facilitating early diagnosis, risk stratification, and improved monitoring across various diseases and their subtypes. In this study, we collected plasma proteomes from over 3000 patients, representing 23 distinct diseases, encompassing a total of 1462 proteins. Based on histological knowledge, we developed a two-stage hierarchical multi-disease classifier and applied it to perform multi-disease classification on the collected proteomic data. Our results demonstrate that this empirically guided two-stage hierarchical multi-disease classifier outperforms traditional machine learning algorithms in terms of prediction performance, showing better balance and more meaningful feature selections. This finding highlights the positive role that domain expertise can play in machine learning-based disease detection, and underscores the potential of plasma proteomics for multi-disease screening.

Background: Machine-learning models based on tissue transcriptomic data are powerful tools for disease classification. However, their clinical adoption is limited by the invasive nature of tissue sampling. Furthermore, transcriptomic datasets are often affected by batch effects and gene-level noise, which compromise model generalizability across platforms and clinical cohorts.

Methods: We developed WBT-DC (Whole Blood Transcriptomics–based Disease Classification), a computational pipeline designed to overcome these challenges. WBT-DC integrates rank-based feature extraction to mitigate batch effects with an ensemble machine-learning framework that incorporates cross-validation and hyperparameter optimization. Its performance was systematically evaluated across five independent cohorts involving 2,164 participants and three disease contexts: Crohn’s disease (CD), ulcerative colitis (UC), and amyotrophic lateral sclerosis (ALS). We tested the model’s robustness across RNA-sequencing and microarray platforms. Additionally, an internal rheumatoid arthritis (RA) cohort ( n  = 165) was utilized for real-world prospective validation.

Results: WBT-DC demonstrated high accuracy, achieving ROC–AUC values of 0.90–0.94 in independent datasets when training and testing were conducted on the same platform. In cross-platform evaluations, the pipeline maintained robust performance with ROC–AUC values ranging from 0.71 to 0.84, consistently outperforming conventional gene expression-based models. In the RA validation cohort, WBT-DC achieved an ROC–AUC of 0.81, supporting its applicability in a real-world clinical setting.

Conclusions: WBT-DC provides a robust, non-invasive, and platform-agnostic framework for disease classification using whole-blood transcriptomics. By effectively addressing batch effects and platform variability, this pipeline offers a scalable solution for translating systems-level transcriptomic insights into applications.

The human blood proteome provides a holistic readout of health states through the assessment of thousands of circulating proteins. In this study, we present a pan-disease resource to enable the study of diverse disease phenotypes within a harmonized proteomics dataset. By profiling protein concentrations across 59 diseases and healthy cohorts, we identified proteins associated with age, sex, and body mass index, as well as disease-specific signatures. This study highlights shared and distinct protein patterns across conditions, demonstrating the power of a unified proteomics approach to uncover biological insights. The dataset, covering 8262 individuals and up to 5416 proteins, serves as an online resource for exploring disease-specific protein profiles and advancing precision medicine research.

Chronic hepatic injury and inflammation from various causes can lead to fibrosis and cirrhosis, potentially predisposing to hepatocellular carcinoma. The molecular mechanisms underlying fibrosis and its progression remain incompletely understood. Using a proteo-transcriptomics approach, we analyze liver and plasma samples from 330 individuals, including 40 healthy individuals and 290 patients with histologically characterized fibrosis due to chronic viral infection, alcohol consumption, or metabolic dysfunction-associated steatotic liver disease. Our findings reveal dysregulated pathways related to extracellular matrix, immune response, inflammation, and metabolism in advanced fibrosis. We also identify 132 circulating proteins associated with advanced fibrosis, with neurofascin and growth differentiation factor 15 demonstrating superior predictive performance for advanced fibrosis(area under the receiver operating characteristic curve [AUROC] 0.89 [95% confidence interval (CI) 0.81–0.97]) compared to the fibrosis-4 model (AUROC 0.85 [95% CI 0.78–0.93]). These findings provide insights into fibrosis pathogenesis and highlight the potential for more accurate non-invasive diagnosis.

Alzheimer's disease (AD) is a debilitating neurodegenerative disorder with a global impact, yet its pathogenesis remains poorly understood. While age, metabolic abnormalities, and accumulation of neurotoxic substances are potential risk factors for AD, their effects are confounded by other factors. To address this challenge, we first utilized multi-omics data from 87 well phenotyped AD patients and generated plasma proteomics and metabolomics data, as well as gut and saliva metagenomics data to investigate the molecular-level alterations accounting the host-microbiome interactions. Second, we analyzed individual omics data and identified the key parameters involved in the severity of the dementia in AD patients. Next, we employed Artificial Intelligence (AI) based models to predict AD severity based on the significantly altered features identified in each omics analysis. Based on our integrative analysis, we found the clinical relevance of plasma proteins, including SKAP1 and NEFL, plasma metabolites including homovanillate and glutamate, and Paraprevotella clara in gut microbiome in predicting the AD severity. Finally, we validated the predictive power of our AI based models by generating additional multi-omics data from the same group of AD patients by following up for 3 months. Hence, we observed that these results may have important implications for the development of potential diagnostic and therapeutic approaches for AD patients.

Various causes of chronic hepatic injury and inflammation can lead to fibrosis and cirrhosis, potentially predisposing individuals to hepatocellular carcinoma. Despite extensive research, the molecular mechanisms underlying liver fibrosis and its associated progression to cancer remain incompletely understood. In this study, we employed an integrated proteotranscriptomics approach to characterize the molecular pathophysiology of liver fibrosis in both liver and plasma samples from 330 individuals. This cohort included 40 healthy subjects and 290 patients with histologically characterized fibrosis due to chronic viral infection, alcohol consumption, or metabolic-dysfunction associated steatotic liver disease. We demonstrated that pathways related to extracellular matrix alterations, immune response, inflammation, and metabolism are dysregulated in advanced hepatic fibrosis, regardless of the underlying cause. Additionally, our analysis of peritumoral hepatic tissues revealed transcription signatures linked to cell proliferation, survival, and inflammation in hepatocellular carcinoma. Furthermore, we observed extensive remodeling of the plasma proteome linked with severe fibrosis and identified 132 circulating proteomic signatures associated with advanced fibrosis by integrative analysis of plasma proteomics with hepatic transcriptomics. We finally developed predictive models using machine learning to facilitate the non-invasive detection of advanced fibrosis and cirrhosis.