I. Database Overview

Welcome to PWAS Atlas database, a comprehensive repository designed to curate, integrate, analyze, and visualize proteome-wide association study (PWAS) data, linking genetically regulated protein abundance to human complex traits and diseases. PWAS Atlas combines manual literature curation with large-scale uniform computational analyses, both curated from PWAS publications and analyzed from a standard workflow. The database currently integrates over 40,000 curated and computationally-derived protein–trait associations, covering 4,585 genes/proteins, 763 traits and 6 tissues. In addition, PWAS Atlas provides interactive knowledge graphs, gene-centered visualizations, and comparative analysis tool for exploring shared and distinct association patterns across multiple genes/proteins or traits. In summary, PWAS Atlas offers a comprehensive platform for exploring the molecular mechanisms through which genetically regulated protein abundance contributes to disease susceptibility and phenotypic variation, facilitating the prioritization of candidate proteins for mechanistic investigation, biomarker assessment, and therapeutic target validation.

Introduction to proteome-wide association study

Proteome-wide association study integrates protein quantitative trait loci (pQTL) data with genome-wide association study (GWAS) data to identify trait-associated proteins regulated by risk genomic loci. In detail, PWAS first builds protein abundance prediction models for each gene/protein using an external reference panel with genotype data and proteomic data. Then the protein abundance of GWAS samples is imputed with these models, and finally the association between genetically predicted protein abundance and traits is calculated to identify trait-associated proteins. As the primary functional executors of biological processes, proteins offer a direct link between genetic variation and disease mechanisms, thereby facilitating the identification of causal proteins, biomarkers, and therapeutic targets.

II. Data Curation

We included all available proteome-wide association study publications in the database. The curation steps were as follows:

1. Literature search and screening: we searched PubMed using predefined keywords. Publications were included if they met the following criteria:
   1. The study performed PWAS by integrating genetically predicted protein abundance with GWAS summary statistics, rather than solely comparing measured protein abundance between cases and controls.
   2. The study reported explicit protein–trait association statistics, such as P-values, Z-scores, false discovery rate (FDR) values, or significance thresholds.
2. Study curation: we manually curated the study information for each qualified publication, including reported trait, tissue type, pQTL reference panel/dataset, method/software, and ancestry of population.
   1. Association collection: we collected protein–trait associations including significance and size of effect of protein abundance on trait. Furthermore, we integrated genome-level regulatory information about genes/proteins (pQTLs) in the corresponding tissues from the original pQTL datasets to provide more comprehensive regulatory information for each trait.

As of May 11, 2026, this curation process has yielded 15,189 protein–trait associations manually curated from 118 PWAS publications, covering 169 traits, 6 tissues, and 4,109 genes/proteins.

III. Data Analysis Workflow

PWAS Atlas performs proteome-wide association study analysis based on protein abundance prediction models built from five publicly available pQTL reference panels and GWAS summary statistics from the UK Biobank, using the FUSION framework in the current version.

pQTL reference panel collection

We incorporated five major population-scale proteomic reference panels, providing genotype data paired with protein abundance measurements. These panels were selected based on the public availability and accessibility of protein abundance prediction weights:

1. UK Biobank Pharma Proteomics Project (UKB-PPP)
2. ARIC Study
3. INTERVAL Study
4. Banner Sun Health Research Institute
5. Religious Orders Study and Rush Memory and Aging Project (ROS/MAP)

These panels are used to build protein abundance prediction models for each gene/protein for downstream association analysis.

GWAS data collection

We obtained GWAS summary statistics from the UK Biobank, covering 841 biologically relevant datasets. The European-ancestry reference panel from the 1000 Genomes Project was used to estimate linkage disequilibrium.

Data preprocessing and standardization

1. Standardize gene and protein identifiers utilizing Ensembl and UniProt references.
2. Obtain protein abundance prediction weights/models from the five publicly available pQTL reference panels.
3. Impute genetically predicted protein abundance into GWAS samples and calculate the association between predicted protein abundance and each trait.

Software

We used FUSION to perform PWAS analysis. FUSION is a framework that integrates reference-panel-based molecular prediction models with GWAS summary statistics to identify trait-associated genes/proteins.

Through this uniform analytical workflow, PWAS Atlas systematically analyzed 666 traits and identified 25,420 significant protein–trait associations involving 2,252 genes/proteins.

IV. Trait Ontology

PWAS Atlas standardizes reported traits using multiple biomedical and life science ontologies, including the Experimental Factor Ontology (EFO), Mondo Disease Ontology (MONDO), Human Disease Ontology (DOID), Human Phenotype Ontology (HPO), Medical Subject Headings (MeSH), and SNOMED CT, with additional resources incorporated when necessary. Each mapped trait is assigned a preferred label, ontology identifier, definition, synonyms, and external cross-references. Both the originally reported trait and its mapped standardized concept are retained to ensure traceability.

V. Gene/Protein Annotation and Standardization

PWAS Atlas standardizes genes using Ensembl release 116 based on the GRCh38 genome assembly and maps proteins to UniProt accession and name fields. Gene/protein entries include standardized identifiers, functional annotations, and synonyms. Gene-level molecular pathway and therapeutic target annotations are further integrated from Gene Ontology (GO) and ChEMBL, with clinical-stage or approved drugs associated with the corresponding gene/protein targets included.

VI. Knowledge Graph

Introduction

We have developed an interactive knowledge graph to facilitate understanding of the complex relationships among genes, proteins, tissues and traits. This graph establishes a unified framework linking genes, proteins, tissues and traits, and is updated as new data becomes available, ensuring a current view of the evolving landscape of PWAS associations.

Implementation

The knowledge graph provides the following central node types:

1. Choose a gene/protein as the central node: for each available gene/protein, the knowledge graph displays gene/protein-tissue-trait layers of association.
2. Choose a trait as the central node: for each available trait, the knowledge graph displays trait-gene/protein-tissue layers of association.

VII. Database Usage

PWAS Atlas consists of five major functional modules:

Browse Module

This module consists of four categories: Gene, Trait, Publication, and Dataset. An overview table is provided for each category, allowing users to conveniently explore genes/proteins, traits, curated publications, datasets, and their inner associations. Detailed pages further provide standardized annotations, association evidence, and links to related records, while statistical summaries offer an overview of the data distribution and content represented in the database.

Search Module

1. Quick Search: empowers users to swiftly search for keywords pertaining to genes, proteins, traits, or publications directly on the home page.
2. Advanced Search: for users seeking more refined queries, an advanced search option is accessible through a dedicated page, enabling selection by gene, trait, and tissue to facilitate precise identification of specific protein–trait associations.

Knowledge Graph

The Knowledge Graph module enables interactive exploration of PWAS associations. Users can select a gene/protein or trait as the central node and visualize its associated relationships across tissues. Associations can be filtered by tissue, effect direction, evidence source, and data origin. The graph layout can be interactively adjusted and exported, and a complementary Sankey diagram provides an overview of association patterns among traits, tissues, and genes/proteins.

Comparative Analysis

The Comparative Analysis module supports gene- or trait-centered comparisons of two to five genes/proteins or traits. The “Shared and Unique Associations” view highlights overlapping and unique associations, while the “Gene-Trait Association Matrix” summarizes association patterns together with contextual information such as tissue and supporting evidence. These functions facilitate systematic comparison of shared and distinct PWAS association patterns.

Download Module

The database provides a repository of downloadable resources, ensuring that users have access to information including curated literature records and standardized annotations, protein prediction reference panel data, GWAS project datasets, summary information for traits, genes, and proteins, and significant protein–trait association data, for secondary analysis and reproducible research.