ProteoNexus

[Website](https://www.proteonexus.com/) [Data](https://doi.org/10.57760/sciencedb.26857)

ProteoNexus characterizes genetic architecture, estimates mediation effects, and constructs and evaluates prediction models of plasma proteome.

This repository contains the official analysis pipeline for ProteoNexus. We provide the source code for pQTL analysis, mediation effect estimation, prediction model training, prediction execution, and data integrity validation tools.

IMPORTANT NOTE This repository contains the backend analysis code and scripts only. It does not include the source code for the user-facing web server, which is available at https://www.proteonexus.com/.

🌐 Web Server

For interactive exploration, visualization, and analysis, please visit the ProteoNexus web platform:

https://www.proteonexus.com/

The platform is freely available and does not require registration.

✨ Key Features

ProteoNexus is a comprehensive platform designed to decode the complex relationships between exposures, the plasma proteome, and disease incidence using data from the UK Biobank (application ID: 144904).

🧬 Genetic Architecture Analysis

• Estimates pQTL summary statistics using GEMMA, adjusting for covariates (age, sex, top 18 genetic PCs).
• Performs fine-mapping analysis with SuSiE to identify probable causal SNPs (PIP).
• Provides interactive visualizations, including Manhattan plots and QQ-plots, on the web server.

🔗 Mediation Effect Estimation

• Systematically investigates the mediating role of 2,919 plasma proteins in three distinct causal pathways:
  1. M-P-D: Measurement → Protein → Disease
  2. E-P-D: Environment → Protein → Disease
  3. G-P-D: SNP → Protein → Disease
• Identifies thousands of significant mediation pathways, ensuring that protein expression was measured before disease incidence.

📈 High-Performance Prediction Models

• Constructs and evaluates optimized prediction models for 57 incident diseases.
• Optimizes four machine learning models using a Tree-structured Parzen Estimator (TPE) to maximize AUC:
  • Penalized Logistic Regression (PLR)
  • XGBoost
  • LightGBM
  • Multi-Layer Perceptron (MLP)
• The web server provides an interface to predict disease probability using our pre-trained models on user-provided proteomic data.

🚻 Sex-Specific Analysis

• Implements a complete sex-specific pipeline for genetic architecture, mediation analysis, and prediction modeling to uncover sex-differentiated biological mechanisms.

💻 Repository Content

This repository includes the core computational scripts for:

• pQTL Analysis: Scripts for running GEMMA and SuSiE.
• Mediation Analysis: R scripts utilizing the medflex package.
• Model Training: Python scripts for hyperparameter optimization of PLR, XGBoost, LightGBM, and MLP models.
• Prediction: Scripts to apply the trained models to new data.
• Data Integrity: Tools for quality control and data validation.

📊 Data Availability

The genome-wide pQTL summary statistics generated from 33,325 European participants for 2,919 plasma proteins are publicly available at the Science Data Bank.

• Citation: Kaixin Shao, Peng Huang, Sheng Yang (2025) . Summary statistics for pQTL in UKB-PPP. V1. Science Data Bank.
• DOI: https://doi.org/10.57760/sciencedb.26857

📜 Citation

If you use the ProteoNexus platform or the code from this repository in your research, please cite:

Shao K, Luo Z, Huang P, Yang S. ProteoNexus: an integrative database to characterize genetic architecture, estimate mediation effects, and construct and evaluate prediction models of the plasma proteome. Nucleic Acids Res. doi: 10.1093/nar/gkaf1058.

🔗 Related Resources from Our Group

GWAShug - Resource for shared genetic basis

Cao C, Tian M, Li Z, Zhu W, Huang P, Yang S. GWAShug: a comprehensive platform for decoding the shared genetic basis between complex traits based on summary statistics. Nucleic Acids Res. 2025 Jan 6;53(D1):D1006-D1015. doi: 10.1093/nar/gkae873.

PGS-Depot - Resource for polygenic score

Cao C, Zhang S, Wang J, Tian M, Ji X, Huang D, Yang S, Gu N. PGS-Depot: a comprehensive resource for polygenic scores constructed by summary statistics based methods. Nucleic Acids Res. 2024 Jan 5;52(D1):D963-D971. doi: 10.1093/nar/gkad1029.

SRT-Server - Server for Spatial Resolved Transcriptomics Data Analysis and visualization

Yang S, Zhou X. SRT-Server: powering the analysis of spatial transcriptomic data. Genome Med. 2024 Jan 26;16(1):18. doi: 10.1186/s13073-024-01288-6.

PGSFusion - Server for polygenic score construction

Yang S, Ye X, Ji X, Li Z, Tian M, Huang P, Cao C. PGSFusion streamlines polygenic score construction and epidemiological applications in biobank-scale cohorts. Genome Med. 2025 Jul 14;17(1):77. doi: 10.1186/s13073-025-01505-w.