Genomic atlas of 7,000 plasma proteins and their associations with diseases and traits in East Asian populations

This study presents a large-scale proteogenomic atlas of 7,289 plasma proteins in 3,965 Chinese adults, identifying thousands of protein-phenotype associations and causal links that reveal ancestry-specific genetic insights and biological networks, particularly in lipid metabolism and cardiovascular disease, which are often missed in European-centric research.

The Gist

The Big Picture: A New Map for a Forgotten Neighborhood

Imagine the human body as a massive, bustling city. For years, scientists have been drawing detailed maps of this city, but they have mostly focused on one specific neighborhood: people of European ancestry. They know which streets (genes) lead to which buildings (proteins) and how those buildings affect the city's traffic (diseases).

This paper says, "Wait a minute, we are missing a huge part of the map!" The researchers decided to draw a detailed map of the same city, but this time for East Asian populations (specifically Chinese adults). They wanted to see if the streets and buildings work the same way here, or if there are unique roads and landmarks that only exist in this part of the world.

The Tools: A High-Tech Scanner and a Genetic Library

To do this, the team used two main tools:

1. The Genetic Library (DNA): They looked at the DNA of nearly 4,000 Chinese adults. Think of DNA as the instruction manual for building the city.
2. The High-Tech Scanner (Proteomics): They used a powerful machine called SomaScan to take a snapshot of the city's current state. This machine measured the levels of about 7,000 different proteins in the blood. Proteins are like the workers, machines, and messengers keeping the city running.

The Discovery: Finding the "Switches" (pQTLs)

The researchers asked a simple question: Which parts of the instruction manual (DNA) control how many workers (proteins) are on the job?

They found 3,212 "switches" (called pQTLs).

• The "Local Switches" (Cis-pQTLs): About 1,000 of these switches are right next to the building they control. It's like a light switch located directly inside a specific factory that controls how many workers that factory produces.
• The "Remote Switches" (Trans-pQTLs): The other 2,000 switches are far away. It's like a control tower in a different city that sends a signal to change the number of workers in a factory miles away.

Key Finding: Most of these switches work very similarly to the ones found in European studies. If a switch turns a protein "up" in Europe, it usually turns it "up" in China too. However, the researchers also found some unique switches that only exist in East Asians. If you only looked at the European map, you would miss these entirely.

The Connections: How Proteins Talk to Diseases

Once they had their map of switches and proteins, they connected it to a massive database of health records (like a city logbook) containing information on hundreds of diseases and traits (like blood pressure, diabetes, or heart disease).

They used a method called Mendelian Randomization (think of it as a "genetic detective") to figure out cause and effect. Instead of just seeing that two things happen together, they used the genetic switches to prove that changing the protein causes a change in the disease risk.

What they found:

• They identified nearly 8,000 strong connections between specific proteins and specific health traits.
• They narrowed this down to 645 high-confidence pairs where the evidence is very strong.
• The "Star" Discovery: They found a specific protein called APOF (Apolipoprotein F). Their genetic detective work showed that having higher levels of APOF is linked to a lower risk of liver cancer. This is a new clue that wasn't clearly seen before. It suggests APOF might be a protective shield for the liver.

The "City Network" Effect

The researchers noticed that the city doesn't work in isolation. Changing one protein often ripples out and changes many others.

• They found "Hub Proteins" (like the APOE protein) that act like major traffic controllers. If you tweak the switch for APOE, it changes the levels of over 100 other proteins in the blood.
• These hubs are heavily involved in lipid (fat) metabolism and heart health, showing that these systems are deeply interconnected.

Why This Matters (According to the Paper)

The paper concludes that this new map is a vital resource.

1. It fills a gap: It proves that studying only European populations leaves out unique genetic clues found in East Asians.
2. It validates the tools: It shows that the high-tech scanners work well across different populations, giving scientists confidence to use them globally.
3. It offers new leads: By finding these unique switches and connections (like the APOF and liver cancer link), the study provides a "cheat sheet" for future scientists to figure out how to treat diseases and develop new drugs specifically tailored to these populations.

In short, the researchers built a massive, detailed instruction manual for the East Asian proteome, showing us which genetic switches control our health and revealing that some of the most important switches were previously hidden from view.

Technical Summary

Problem Statement
Proteogenomic studies integrating genetic, molecular, and phenotypic data have transformed target discovery but remain heavily biased toward European populations. While large-scale proteomic studies using platforms like Olink and SomaScan have been conducted in European and Icelandic cohorts, systematic cross-ancestry evaluation of SomaScan pQTL replicability in East Asian populations is limited. Furthermore, previous studies utilized assays with fewer aptamers (~5,000), whereas newer technology allows for the measurement of up to 11,000 aptamers. There is a critical need to characterize the genetic architecture of the plasma proteome in non-European populations to identify ancestry-specific variants and establish a comprehensive atlas of protein-phenotype relationships for drug-target prioritization.

Methodology
The study utilized data from 3,965 Chinese adults from the China Kadoorie Biobank (CKB), comprising incident ischemic heart disease (IHD) cases and a sub-cohort.

• Proteomics: Plasma proteins were profiled using the SomaScan v4.1 assay, measuring 7,596 aptamers (targeting ~7,289 human proteins). Data were normalized using adaptive normalization with maximum likelihood (ANML).
• Genetics: Genome-wide association studies (GWAS) were performed on ~11 million imputed variants. The study identified protein quantitative trait loci (pQTLs) at a Bonferroni-corrected significance threshold.
• Analysis Framework:
  • pQTL Discovery: Identified cis- and trans-pQTLs and performed Bayesian fine-mapping (SuSIE) to define credible sets of causal variants.
  • Cross-Ancestry/Platform Comparison: Compared results with the deCODE (Icelandic, SomaScan) and UK Biobank Pharma Proteomics Project (UKB-PPP, European, Olink) datasets.
  • Phenome-Wide Association Study (PheWAS): Tested 3,107 lead pQTLs against 225 East Asian traits derived from BioBank Japan (BBJ), the Korean Genome and Epidemiology Study (KoGES), and the Taiwan Biobank (TWB).
  • Causal Inference: Performed colocalization analyses (requiring PP.H4 > 0.8) and Mendelian Randomization (MR) using cis-pQTL instruments (Wald ratio and GSMR) to prioritize causal protein-phenotype associations.
  • Network Analysis: Constructed protein-protein regulatory networks based on cis-trans effects.

Key Results

• pQTL Discovery: The study identified 3,212 pQTL associations, including 1,282 cis-pQTLs and 1,825 trans-pQTLs. These associations mapped to 715 non-overlapping genomic regions. Notably, 1,092 proteins had a cis-pQTL, and 179 aptamers showed strong trans-pQTLs without corresponding cis-pQTLs.
• Replicability and Ancestry: 87% of lead pQTLs identified in this East Asian cohort were replicated in the deCODE SomaScan study with concordant effect directions. Effect sizes were highly correlated (Pearson r = 0.93 for cis-pQTLs). However, replication rates were lower when comparing SomaScan (CKB) to Olink (UKB-PPP), particularly for trans-pQTLs, highlighting platform-specific differences.
• Fine-Mapping: Fine-mapping resolved 666 cis-pQTLs and 1,471 trans-pQTLs to single-SNP resolution. Approximately 30% of cis-pQTLs involved protein-altering variants (PAVs), such as missense mutations or stop-gained variants.
• Protein Networks: A regulatory network was identified where 305 cis-proteins exerted downstream effects on 1,018 trans-proteins. Prominent hubs included apolipoproteins (APOE, APOC3, APOA5), complement components, and coagulation factors.
• Protein-Phenotype Associations:
  • PheWAS: Identified 17,980 protein-trait associations (FDR < 0.05), with a strong clustering of cardiometabolic traits (e.g., lipid levels, coronary artery disease) and blood pressure traits.
  • Colocalization: Found 7,936 protein-trait pairs with strong colocalization support (PP.H4 > 0.8). Most shared signals (91%) were in trans regions.
  • Mendelian Randomization: Prioritized 1,975 protein-trait associations using cis-pQTL instruments. Of these, 645 pairs were supported by both colocalization and causal inference.
  • Novel Findings: The study identified an association between apolipoprotein F (APOF) and hepatic cancer, where higher genetically predicted APOF levels were associated with reduced risk (OR ~0.78-0.79). This finding aligns with prior tissue-based studies showing APOF downregulation in hepatocellular carcinoma.
  • Ancestry-Specific Insights: The study identified ancestry-specific pQTLs that contributed to associations undetectable in European studies alone, emphasizing the value of diverse ancestry data.

Significance and Claims
The authors claim this study provides the first large-scale proteogenomic atlas in a non-European population, expanding the global understanding of the plasma proteome's genetic architecture. By integrating ~7,000 proteins with East Asian phenotypes, the study establishes a publicly valuable resource of genetically anchored protein-phenotype relationships. The authors assert that this atlas facilitates:

1. Target Prioritization: Providing a foundation for drug-target prioritization and risk-benefit assessment.
2. Ancestry-Inclusive Science: Demonstrating that ancestry-specific instruments can reveal associations missed in European-centric studies.
3. Biological Insight: Organizing associations into coherent biological networks, particularly involving lipid metabolism and cardiovascular disease.

The paper concludes that while the resource is foundational, further expansions into rare-variant analyses and better-powered non-European GWAS are required to fully resolve the proteogenomic architecture underlying human disease. The authors maintain a modest tone regarding clinical application, noting that associations require functional validation and that the study is a preprint not yet certified by peer review.