Multi-Phenotype Approach Highlights Increased Power of cPCA When Identifying CKD-Related Loci

Study results published in PLOS Genetics demonstrate how a multi-phenotype approach highlights the increased power of combinatorial principal component analysis (cPCA) in identifying chronic kidney disease (CKD)-related loci. The authors of the study wrote that, by integrating multiple measurements, the findings offer a clearer understanding of CKD’s genetic basis, helping create a foundation to develop similar approaches for other complex diseases. Such innovations can lead to prevention and treatment.¹

Image credit: Crystal light | stock.adobe.com

CKD is a multifactorial condition that is driven by diverse etiologies—such as hypertension, diabetes, infections, and lifestyle factors—and leads to a gradual loss of kidney function. Although previous genome-wide association studies (GWAS) have identified numerous genetic loci that are associated with CKD, there is still a large portion of genetic basis that are unknown.¹

“[CKD’s] complexity makes it difficult to capture the full picture of kidney health using only a single biomarker, such as the commonly used estimated glomerular filtration rate (eGFR), to assess kidney function, and therefore may miss important details. This incomplete understanding of the genetics for different CKD subtypes has hindered the identification of drug targets to treat the various subtypes,” study author Kim Ngan Tran, PhD, Queensland University of Technology School of Biomedical Sciences, Centre for Genomics and Personalised Health, explained in a news release.²

To address this the investigators implemented a novel multi-phenotype approach, cPCA, to better understand CKD’s complex genetic architecture. Drawing from 337,112 individuals from a UK Biobank dataset, the investigators analyzed 21 different CKD-related phenotypes, generating over 2 million composite phenotypes (CPs) through cPCA.¹

The investigators explained that the primary objectives of cPCA were to identify optimal combinations of biomarkers that collectively enhance disease classification, outperforming individual biomarkers alone, and to identify genetic loci that are linked to target disease by utilizing multi-trait GWAS approaches. Additionally, cPCA shares similarities with multivariate GWAS as well as other joint analyses of multiple traits, including PCA-based methods; however, the authors explained that unlike traditional multivariate methods that require a predefined set of traits, cPCA is able to systematically explore and select the optimal biomarker combinations, making it distinct and more flexible.¹

Among the identified CPs, nearly 50,000 were observed to have a significantly higher classification power for clinical CKD compared with individual biomarkers. The top-ranked CP—which was a combination of albumin, cystatin C, eGFR, gamma-glutamyltransferase, hemoglobulin A_1c, low-density lipoprotein, and microalbuminuria—was observed to achieve an area under the curve (AUC) of about 0.878 (95% CI, 0.873–0.882), which significantly outperformed eGFR alone (AUC: 0.830; 95% CI, 0.825–0.835). Additionally, genetic association analysis of the nearly 50,000 high-performing CPs had identified all major eGFR-associated loci. However, SH2B3 locus rs3184504, which is a loss-of-function variant, which was uniquely identified in CPs (p = 3.1 x 10^-56) but not in eGFR within the same sample size. SH2B3 locus had also showed strong evidence of colocalization with eGFR, supporting its role within kidney function.¹

“These more informative traits enabled the discovery of genetic signals that traditional methods had missed. For example, we identified a variant in the SH2B3 gene, which was previously detected only in large-scale studies involving more than a million individuals, as significantly associated with kidney function using our composite traits. This locus was not detected using traditional single biomarkers like eGFR in our dataset but emerged consistently across many of our high-performing composite phenotypes, highlighting the power of our approach to uncover important genetic associations even in moderately sized cohorts,” Tran said.²

One limitation that the authors noted was the cPCA’s difficulty differentiating between biomarkers and casual genes, despite its strong discriminatory power. They hypothesized that this may have stemmed from the inclusion of biomarkers which reflect CKD status, even if it is not directly involved in the disease’s underlying mechanisms. Further, authors emphasized that these results highlight the capabilities of the multi-phenotype cPCA approach in the understanding of CKD’s genetic basis, and can potentially be applied to other complex diseases.¹

“This study demonstrates the value of an exhaustive yet interpretable multi-phenotype approach to understanding the genetics of CKD and could be applied across large biobanks, smaller deeply phenotyped cohorts, and potentially extended to uncover genetic signals in other complex diseases,” Tran concluded in the news release.²

REFERENCES

1. Tran KN, Sutherland HG, Mallett AJ, Griffiths LR, Lea RA. New composite phenotypes enhance chronic kidney disease classification and genetic associations. PLOS Genetics. Published 2025, May 23. doi:10.1371/journal.pgen.1011718

2. Queensland University of Technology. Study finds better way to detect genetic factors in chronic kidney disease. News release. June 10, 2025. Accessed June 12, 2025. https://www.eurekalert.org/news-releases/1087050