Patient-specific logical models replicate phenotype responses to psoriatic and anti-psoriatic stimuli

Abstract

AbstractPsoriasis is a dermatologic disease that affects 2% of the world population. Psoriasis is characterized by chronic inflammation and aberrant behavior of keratinocytes, which display increased levels of proliferation, and decreased differentiation and apoptosis. Stimulation of keratinocytes by psoriatic cytokines leads to the increased production of immunostimulatory ligands that further attract immune cells and amplify inflammatory responses. Psoriasis can have severe, moderate, or mild outcomes and while these severity levels demand custom medical treatment schemes, assigning an effective treatment to patients with moderate or severe disease is a demanding task.The varied responses of patients to treatments highlight a large disease complexity, demanding that new ways to analyze and integrate patients’ molecular profiles are developed to design patient-specific therapies. We have used gene expression values from psoriasis biopsies to separate patients into two clusters, each with distinct expression profiles, but nevertheless not correlating with any of the available clinical data, such as disease severity. When using these gene expression levels in logical model simulations these data became highly descriptive of patient-specific phenotype characteristics. Starting from a psoriatic keratinocyte model that we published recently, we added additional pathways highlighted by a differential gene expression analysis between the subgroups. This included components from the Interleukin-1 family, IFN-alpha/beta and IL-6 signaling pathways. Model personalization was performed by using patient gene expression levels in model configurations, exploiting the PROFILE pipeline. Personalized simulations revealed that the two patient clusters represent more innate immunity-driven, highly inflammatory phenotypes and adaptive immunity-driven, chronic phenotypes, respectively. The model was also able to finely capture differences between responses in patients with a known disease severity. A treatment response analysis among the patient cohort predicted differential responses to the inhibition of psoriatic stimuli, with IL-17, TNFα and PGE2 inhibition reducing proliferation and inflammatory phenotypes. Alternative treatment with PGE2 or TNFα inhibition instead of IL-17 was suggested for patients with high NF-κB activity and prosurvival factors, such as CREB1. With this project, we aim to highlight the value of combining omics data with logical modeling for the detection of ‘emergent’ phenotypes and for gaining disease knowledge on the individual patient level.