Identifying protein biomarkers and therapeutic targets in psoriasis through integrative genomic, proteomic and transcriptomic analysis

Integrative multi-omics analysis identifies 12 high-priority protein biomarkers and therapeutic targets in psoriasis

Psoriasis is a chronic, immune-mediated inflammatory skin disorder. It affects over 125 million people globally. Modern biologics—drugs designed to target specific proteins in the immune system—have transformed treatment. However, they are not a universal cure. Many patients face inadequate efficacy or adverse side effects. This happens because the molecular drivers of the disease vary between individuals.

Researchers seek reliable protein biomarkers in the blood. These could predict disease risk or guide therapy. But identifying truly causal proteins is difficult. It is hard to tell if a protein drives inflammation or is simply a bystander. This paper addresses that challenge. The authors combine massive genetic datasets with proteomic and transcriptomic layers. Their goal is to move beyond correlation toward a prioritized list of high-confidence therapeutic targets.

The bottleneck of causal inference in proteomics

Current research often relies on Genome-Wide Association Studies (GWAS). These identify genetic variants associated with a disease. However, GWAS usually points to genomic "neighborhoods" rather than specific molecules. A researcher might find a signal on a chromosome. They still won't know if a specific protein or a regulatory element causes the disease.

Distinguishing between cause and consequence is a major hurdle. During intense inflammation, many proteins increase in concentration. Does a protein drive the inflammation or is it simply produced by the skin in response to it? Previous studies were limited by small sample sizes or narrow protein coverage. They also lacked frameworks to filter out "linkage-driven" associations. These occur when a genetic signal looks like it affects a protein only because it is physically near the actual causal gene.

A multi-layered filtration pipeline

To overcome these hurdles, the authors used a multi-stage prioritization framework. They layered several methodologies to ensure only the most robust candidates survived.

1. Two-sample Mendelian Randomisation (MR): The researchers used protein quantitative trait loci (pQTLs)—genetic variants that influence protein levels—as "natural experiments." Because these variants are assigned at conception, they help infer if changing protein abundance actually causes psoriasis risk. This helps bypass the problem of reverse causation.
2. Colocalisation Analysis: The authors used pairwise conditional colocalisation (PWCoCo) to ensure the genetic signals were shared. They looked for a high posterior probability ($PP.H4 > 0.80$). This value indicates a high likelihood that a single shared variant underlies both the protein and the disease [Figure 2C-2D].
3. SMR and HEIDI Testing: The authors applied Summary-data–based MR (SMR) and Heterogeneity in Dependent Instruments (HEIDI) testing. This uses expression quantitative trait loci (eQTL)—variants that affect mRNA levels—to exclude associations driven by nearby genetic neighbors (linkage disequilibrium).
4. Functional and Druggability Profiling: Finally, they mapped candidates onto biological networks. They checked how they interact with immune pathways. They also used the "Genome for REPositioning" (GREP) to see if existing, approved drugs could be repurposed.

This workflow is shown in the study design flowchart .

From 78 proteins to 12 high-priority targets

The scale of this data integration is significant. The authors analyzed 2,027 and 1,836 plasma proteins from the UK Biobank Pharma Proteomics Project and deCODE genetics. They compared these against a psoriasis GWAS meta-analysis. This meta-analysis included 36,466 cases and 458,078 controls.

The initial MR analysis identified 78 unique proteins with genetically predicted associations to psoriasis [Figure 2A-2B]. After applying the stringent colocalisation filter, 27 proteins remained. Through the final SMR-HEIDI triangulation, the authors identified 12 "Tier 1" proteins: STX4, FLT3, NFKB1, IL18, PRSS53, SPAG1, SGSH, PLAT, RALB, TNFSF11, SPHK2, and STAT3.

The biological connectivity of these targets is striking. Network analysis showed that STAT3 and NFKB1 act as central "hub nodes." This means they have many physical and functional connections to other inflammatory genes .

The study also found that these targets are enriched in existing drug classes. Specifically, they showed a 3.6-fold enrichment in "antineoplastic and immunomodulating agents" (ATC group L). They also showed a 4.5-fold enrichment in drugs targeting "blood and blood-forming organs" (ATC group B) [Table 2].

Biological validation via single-cell sequencing

To confirm these genetic predictions, the authors used single-cell RNA sequencing (scRNA-seq). This technology lets researchers look at the gene expression of individual cells. This avoids the error of averaging signals across an entire tissue sample.

They examined healthy, non-lesional, and lesional psoriatic skin. They found that the expression of these prioritised genes was highly cell-type specific. For example, NFKB1 was enriched in dendritic cells and macrophages in lesional skin. Crucially, the authors looked at how these genes responded to IL-23 blockade (using the drug risankizumab). They saw that some genes like SPAG1 were downregulated in keratinocytes (skin cells) after treatment. Other genes like STAT3 and FLT3 were upregulated in immune and stromal cells .

This shows the targets are dynamically involved in tissue remodeling during treatment.

Limitations and the road ahead

Several caveats remain. First, the genetic data used for MR were mostly of European ancestry. This may limit how well these findings apply to global populations. Second, the study focused on common genetic variants (appearing in $\ge$ 1% of the population). Therefore, rare mutations with potent effects were not captured.

Additionally, the study notes difficulty in certain areas. At the PRSS53 locus on chromosome 17, current methods cannot always resolve the exact causal gene. This happens when multiple candidate genes are very closely linked.

The Verdict

This study provides a high-resolution roadmap for psoriasis drug discovery. The authors moved away from simple association. Instead, they used a multi-layered causal framework to distill thousands of possibilities into 12 high-confidence candidates. The "druggability" analysis suggests some targets could be reached by repurposing oncology drugs. However, the safety of such drugs for chronic skin disease requires further study. For researchers seeking the next generation of precision medicines, this Tier 1 list offers a robust starting point.

Figures from the paper