Multi-Omics Molecular Profiling Enables Rapid Diagnosis of Erythrodermic Skin Diseases

Erythroderma is an acute and potentially life-threatening inflammatory condition characterized by redness and scaling of > 90% of the skin. Its treatment is challenging because various underlying skin diseases can cause erythroderma and are difficult to distinguish. Here, we performed in-depth proteomics and transcriptomics analyses of skin from 96 patients with erythroderma caused by five different diseases, including pityriasis rubra pilaris, psoriasis, atopic dermatitis, cutaneous T-cell lymphoma, and drug-induced maculopapular rash. High-throughput workflows enabled in-depth molecular profiling, identifying over 9,300 proteins and 17,200 protein coding transcripts, revealing distinct molecular signatures for each disease. The proteome showed elevated expression of type 2 immunity associated Charcot-Leyden crystal in skin of atopic dermatitis, potentially contributing to NLRP3-driven chronic inflammation in this disease. Complementary transcriptomic analysis demonstrated selective upregulation of IL17C in pityriasis rubra pilaris, strongly correlating with increased IL1 family cytokine expression. Interestingly, only a subset of these patients expressed this IL17C-IL1 signature, suggesting treatment-relevant disease endotypes. Through multi-omics integration, we uncovered disease-specific molecular signatures consistently altered at both protein and transcript levels. In particular, we identified elevated expression of T-cell regulator RASAL3 in cutaneous T-cell lymphoma, which has not been explored in its pathogenesis so far. To translate these molecular profiles into clinical utility, we expanded our adaptive machine-learning algorithm (ADAPT-Mx) for tissue based-disease classification. This achieved 76.6% diagnostic accuracy, substantially outperforming combined conventional clinical and histopathological methods (59.5%). This study provides a template for precision diagnostics in erythroderma and demonstrates the clinical potential of multi-omic profiling in severe inflammatory skin diseases.