📄 genetic and genomic medicine

Copy number variant association analysis in 94,730 Chinese adults reveals loci influencing anthropometric and cardiometabolic traits

A genome-wide association study of 94,730 Chinese adults from the China Kadoorie Biobank identified 19 independent copy number variant associations across 15 loci influencing anthropometric and cardiometabolic traits, including novel findings and replicated dosage-sensitive regions, thereby expanding the genetic understanding of these traits in East Asian populations.

Original authors: Howard, I., Millwood, I., Morris, S., Lin, K., Avery, D., Yu, C., Lv, J., Sun, D., Pei, P., Li, L., Chen, J., Chen, Z., Walters, R., Bragg, F., Bennett, D.

Published 2026-08-13

📖 4 min read☕ Coffee break read

CC BY 4.0

Original authors: Howard, I., Millwood, I., Morris, S., Lin, K., Avery, D., Yu, C., Lv, J., Sun, D., Pei, P., Li, L., Chen, J., Chen, Z., Walters, R., Bragg, F., Bennett, D.

Original paper licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). ⚕️ This is an AI-generated explanation of a preprint that has not been peer-reviewed. It is not medical advice. Do not make health decisions based on this content. Read full disclaimer

Imagine your body's genetic code as a massive, intricate instruction manual for building and running a human being. For a long time, scientists thought the most important typos in this manual were single-letter mistakes, like swapping an 'A' for a 'G'. These are called single-nucleotide polymorphisms, or SNPs, and they've taught us a lot about why we have blue eyes or why some people get heart disease. But there's another, much bigger kind of typo that's been hiding in plain sight: Copy Number Variants, or CNVs.

Think of SNPs as a single misspelled word in a sentence. CNVs are more like entire paragraphs being accidentally deleted, or whole chapters being pasted in twice. If a paragraph tells your body how to make a hormone, deleting it might mean you make none at all, while pasting it twice might mean you make too much. These "paragraph typos" can have huge effects on everything from your height to your blood sugar. Until now, most scientists have been reading these genetic instruction manuals mostly in people of European ancestry. It's like trying to understand how a car engine works by only looking at models made in one specific factory. We've been missing out on how these genetic "paragraphs" work in other parts of the world, where the manuals might be written in a slightly different dialect.

This paper is a massive step toward filling that gap. The researchers took a giant leap into the genetic library of nearly 95,000 adults from China, part of the China Kadoorie Biobank. They didn't just look for single-letter typos; they hunted for the big "paragraph" changes—the CNVs—and asked: do these big genetic shifts change how much fat we carry, how high our blood pressure is, or how our bodies handle sugar?

The team found 19 distinct places in the genetic manual where these copy-number changes seem to act like volume knobs for our health. Some of these findings were like finding familiar landmarks in a new city. For instance, they confirmed that a specific region on chromosome 16 acts as a powerful switch for body weight. In this group, having an extra copy of this genetic chunk was like turning the volume down on fat storage, leading to significantly lower body mass index (BMI) and body fat percentage. Having a missing chunk did the opposite, pushing weight up. This confirmed that these "volume knobs" work similarly in Chinese adults as they do in Europeans, even though the average body weight in the study group was lower.

But the real excitement came from the new discoveries—places in the manual no one had noticed before. The researchers found that duplications in a tiny section of chromosome 12 and another on chromosome 17 were linked to lower body fat percentage, but surprisingly, these didn't seem to change a person's overall BMI. It's as if these genetic changes were specifically tuning the "fat meter" without touching the "weight scale." They also spotted a link between a genetic deletion on chromosome 8 and lower random plasma glucose (a measure of blood sugar), and a connection between a genetic change on chromosome 7 and higher diastolic blood pressure.

The study didn't just find new spots; it also showed that the "volume knobs" work in both directions. At the chromosome 16 spot, losing a piece of DNA made people heavier, while gaining a piece made them lighter. This suggests that the amount of genetic material matters a lot, not just the presence or absence of a specific gene.

However, the authors are careful to note that while these links are statistically strong, they are still just associations. They haven't proven exactly how these genetic changes cause the changes in body fat or blood pressure yet. It's like finding that a specific switch is always flipped when the lights are on, but not yet knowing the wiring diagram inside the wall. Also, because the study was done in one specific population, we don't know yet if these exact same switches work the same way in other parts of Asia or the world.

In short, this paper is a treasure map for a part of the genetic world we haven't explored much. It shows us that in Chinese adults, big genetic changes play a significant role in shaping our bodies and health, sometimes in ways we didn't expect. By finding these new "paragraph typos," the researchers are helping us build a more complete picture of human health, one that includes people from all over the globe, not just those from one corner of the world.

Technical Summary: Copy Number Variant Association Analysis in 94,730 Chinese Adults

Problem Statement
Genome-wide association studies (GWAS) have largely relied on common single-nucleotide polymorphisms (SNPs), which explain only a fraction of the phenotypic variance in complex traits. Copy-number variants (CNVs), defined as deletions or duplications of genomic segments typically ≥1,000 base pairs, represent a critical source of genetic variation capable of altering gene dosage and regulatory elements. However, large-scale CNV association studies have been heavily concentrated in European-ancestry populations. Existing studies in East Asian populations have been limited by small sample sizes (often <10,000 individuals), restricting the ability to detect rare CNVs with moderate effect sizes or to perform comprehensive phenome-wide discovery. Furthermore, CNV frequencies and trait architectures vary across ancestries, suggesting that findings from European cohorts may not be transferable. This study addresses the gap in understanding the contribution of structural variation to anthropometric and cardiometabolic traits in a large East Asian population.

Methodology
The study utilized data from the China Kadoorie Biobank (CKB), a prospective cohort of 512,714 individuals. The analysis focused on 94,730 adults (after quality control) with genotypic and phenotypic data.

Genotyping and CNV Calling: Participants were genotyped using two custom Affymetrix Axiom arrays (CKB1 and CKB2) optimized for East Asian ancestry. Raw intensity data were processed to generate normalized log R ratio (LRR) and B-allele frequency (BAF) values. CNVs were called using PennCNV v1.0.5 with a hidden Markov model.
Quality Control (QC): Rigorous QC was applied at multiple levels. Sample-level filters excluded arrays with poor quality control metrics, sex mismatches, or excessive heterozygosity. Probe-level filters removed markers with poor clustering or high missingness. Post-calling filters excluded CNVs overlapping centromeric/telomeric regions or segmental duplications, and samples with high waviness factors or excessive CNV counts. High-confidence CNVs were defined as those with a confidence score ≥20, spanning ≥10 kb, and supported by ≥10 probes.
Association Analysis: The study analyzed 13 quantitative traits: nine anthropometric (e.g., BMI, body fat percentage, height), three cardiovascular (systolic/diastolic blood pressure, resting heart rate), and one glycaemic (random plasma glucose).
- Three association models were employed: a "mirror" model (coding deletions as -1 and duplications as +1), a deletion-only model, and a duplication-only model.
- Linear regression was performed on inverse rank-normal transformed (RINT) phenotypes, adjusting for sex, age, age², study region, array version, and the first 11 genetic principal components.
- Multiple testing correction utilized the SimpleM method to estimate the effective number of independent tests ( $M_{eff}$ ), establishing CNV-wide significance thresholds ( $P \le 8.40 \times 10^{-5}$ for deletions, $P \le 1.33 \times 10^{-6}$ for duplications, and $P \le 9.24 \times 10^{-5}$ for mirror models).
Signal Consolidation: Significant probes were consolidated into Copy-Number Variable Regions (CNVRs) using a linkage disequilibrium (LD) tagging approach ( $r^2 \ge 0.5$ within $\pm 3$ Mbp).

Key Contributions and Results
The study identified 19 independent locus-phenotype associations across 15 unique loci in 94,730 Chinese adults.

Replication of Established Dosage-Sensitive Loci:
- 16p11.2: The study replicated strong bidirectional dosage effects at two distinct intervals within 16p11.2 (BP2-BP3 and BP4-BP5).
  - The BP2-BP3 region (overlapping SH2B1) showed associations with five adiposity traits (BMI, weight, waist/hip circumference, body fat percentage). Duplications were associated with lower adiposity (e.g., $\beta = -0.84$ SD per copy for BMI), while deletions were associated with higher adiposity.
  - The BP4-BP5 region showed similar bidirectional effects specifically on body fat percentage.
- 1q21.1: A standing height association was detected, consistent with previous reports of CNV syndromes affecting growth.
Novel Associations:
- Body Fat Percentage: Two novel loci were identified where duplications were associated with reduced body fat percentage:
  - 12p12.1: Overlapping the BCAT1 gene (involved in branched-chain amino acid metabolism).
  - 17q12: A small region proximal to the HNF1B locus, though the CNVR itself did not overlap the canonical HNF1B interval.
- Diastolic Blood Pressure (DBP): A mirror-model signal at 7p21.1 showed bidirectional effects, with duplications associated with higher DBP and deletions with lower DBP.
- Random Plasma Glucose (RPG):
  - 8p23.1: A deletion signal associated with lower RPG, remaining significant after adjustment for time since last meal.
  - 14q11.2: A mirror-model signal showing bidirectional effects, though caution is advised due to potential somatic V(D)J recombination in the T-cell receptor locus.
- Resting Heart Rate: A deletion signal at 1q12 was identified, which largely overlapped with the 1q21.1 RPG signal, suggesting shared structural events.
CNV Landscape:
- The study characterized the CNV landscape in this cohort, finding that heterozygous deletions and duplications accounted for the majority of calls. Duplications were found to be substantially longer than deletions (median 210 kb vs. 77 kb).
- Recurrent hotspots were observed at known dosage-sensitive regions (e.g., 16p11.2, 22q11.2, 15q11.2) and structurally complex regions like 8p23.1 and the MHC (6p21.3).

Significance and Claims
The authors claim this study represents the largest genome-wide CNV association analysis of quantitative traits in an East Asian population to date. The primary significance lies in:

Ancestral Diversity: It expands the ancestry diversity of CNV association studies beyond the predominantly European cohorts, demonstrating that dosage-sensitive loci (such as 16p11.2) generalize across populations with different adiposity profiles.
Novel Biological Insights: The identification of novel loci (e.g., 12p12.1, 17q12) suggests that structural variation contributes to cardiometabolic trait variation in ways not captured by SNP-based GWAS or previous European-centric CNV studies.
Phenotype Specificity: The study highlights that certain CNVs may influence specific traits (e.g., body fat percentage) without affecting others (e.g., BMI), potentially reflecting population-specific adiposity biology where East Asian populations exhibit higher body fat at equivalent BMIs.
Methodological Robustness: The use of probe-level association testing, array-stratified QC, and sensitivity analyses in unrelated individuals supports the robustness of the reported signals, despite the challenges of rare variant detection.

The paper concludes that these findings support the inclusion of structural variation in the genetic architecture of complex traits and underscore the necessity of diverse population studies to fully characterize disease susceptibility and causal pathways. The authors note that while some established associations (e.g., 22q11.2 for BMI) were not recovered, this likely reflects reduced power in population-based cohorts for highly penetrant CNVs and known frequency differences across ancestries.

Technical Summary: Copy Number Variant Association Analysis in 94,730 Chinese Adults

More like this