Acta Dermato-Venereologica 99-13CompleteContent | Page 16

1258 INVESTIGATIVE REPORT Multicomponent Biomarker Approach Improves the Accuracy of Diagnostic Biomarkers for Psoriasis Vulgaris Ene REIMANN 1# , Freddy LÄTTEKIVI 1# , Maris KEERMANN 2,3 , Kristi ABRAM 2,3 , Sulev KÕKS 4,5 , Külli KINGO 2,3 and Alireza FAZELI 1,6 1 Department of Pathophysiology, 2 Department of Dermatology, University of Tartu, 3 Clinic of Dermatology, Tartu University Hospital, Tartu, Estonia, 4 Centre for Comparative Genomics, Murdoch University, 5 The Perron Institute for Neurological and Translational Science, Perth, Australia and 6 Department of Oncology and Metabolism, The Medical School, University of Sheffield, Sheffield, UK # Both authors contributed equally to this manuscript. Accurate biomarker-based diagnosis of psoriasis vulga­ ris has remained a challenge; no reliable disease-spe­ cific biomarkers have yet been identified. There are se­ veral different chronic inflammatory skin diseases that can present similar clinical and dermoscopy features to psoriasis vulgaris, making accurate diagnosis more dif­ ficult. Both literature-based and data-driven selection of biomarker was conducted to select candidates for a multicomponent biomarker for psoriasis vulgaris. Sup­ port vector machine-based classification models were trained using gene expression data from locally recrui­ ted patients and validated on 7 public datasets, which included gene expression data of other inflammatory skin diseases in addition to psoriasis vulgaris. The re­ sulting accuracy of the best classification model based on the expression levels of 4 genes (IL36G, CCL27, NOS2 and C10orf99) was 96.4%, outperforming clas­ sification based on other marker gene combinations, which were more affected by variability in gene expres­ sion profiles between different datasets and patient groups. This approach has the potential to fill the void of clinically applicable diagnostic biomarkers for pso­ riasis vulgaris and other inflammatory skin diseases. Key words: psoriasis; transcriptome; support vector machine. Accepted Oct 14, 2019: E-published Oct 14, 2019 Acta Derm Venereol 2019; 99: 1258–1265. Corr: Freddy Lättekivi, Department of Pathophysiology, University of Tartu, 14b Ravila Str., EE-50411 Tartu, Estonia. E-mail: [email protected] P soriasis can be a challenging disease to diagnose, as no reliable disease-specific and clinically viable bio- markers have been identified. Psoriasis is one of the most prevalent chronic inflammatory autoimmune diseases of the skin, affecting 2–3% of the population worldwide (1). Clinically, psoriasis can manifest itself in a broad spec- trum of subtypes, the most common of which is psoriasis vulgaris (PsV) (2). PsV can share several similar clinical and molecular features with other chronic inflammatory skin conditions, such as parapsoriasis, lichen planus, pity- riasis rosea, contact eczema, and atopic dermatitis (3–5). All these diseases, however, have different treatment and disease management strategies (6–9), which necessitates accurate and biomarker-based diagnosis. doi: 10.2340/00015555-3337 Acta Derm Venereol 2019; 99: 1258–1265 SIGNIFICANCE This article highlights the issue that previously proposed gene expression-based biomarkers for psoriasis vulgaris are dataset-specific, and therefore perform less accurately in alternative groups of patients. This paper addresses this problem by demonstrating that combining several potential biomarkers into a single multicomponent biomarker results in more accurate classification models, which retain their accuracy across different datasets and patient groups with the inclusion of other inflammatory skin diseases. We be- lieve that this approach has the potential to fill the void of clinically applicable diagnostic biomarkers for psoriasis vulgaris and other inflammatory skin diseases. Currently, the diagnosis of PsV relies mainly on the assessment of visible and dermoscopic symptoms by a clinician (10–13) or, in borderline cases, histological evaluation (2). Disease-specific biomarkers could also improve the diagnosis of PsV overlapping with other papulosquamous disorders and provide more accurate means of quantifying treatment efficacy. Years of high- throughput “omics” research have resulted in multiple potential biomarkers for PsV and other inflammatory skin diseases, yet most of them are universal for inflammation and no single marker has shown to be robustly specific to a given disease (14). The approach of using multiple biomarkers in combi- nation has proven to be more robust and accurate than the standalone measurements of individual biomarkers in the context of several human diseases (15–17). With the advancement of computational methods, multicom- ponent biomarkers offer a solution to overcome interin- dividual variability in the biomarker-based diagnostics of complex heterogeneous diseases (18–20). As such, we hypothesized that using a set of biomarkers in com- bination as a multicomponent biomarker would allow PsV to be distinguished from other inflammatory skin diseases more accurately than any of the individual mar- kers independently. Coupling this with a low-cost gene expression quantification method, such as quantitative real-time PCR (qPCR) could result in a clinically viable tool for the diagnosis of PsV. This is an open access article under the CC BY-NC license. www.medicaljournals.se/acta Journal Compilation © 2019 Acta Dermato-Venereologica.