A prediction model identifying glycolysis signature as therapeutic target for psoriasis

The hyperproliferation featured with upregulated glycolysis is a hallmark of psoriasis. However, molecular difference of keratinocyte glycolysis amongst varied pathologic states in psoriasis remain elusive.

To characterize glycolysis status of psoriatic skin and assess the potential of glycolysis score for therapeutic decision.

We analyzed 345414 cells collected from different cohorts of single-cell RNA seq database. A new method, Scissor, was used to integrate the phenotypes in GSE11903 to guide single-cell data analysis, allowing identification of responder subpopulations. AUCell algorithm was performed to evaluate the glycolysis status of single cell. Glycolysis signature was used for further ordering in trajectory analysis. The signature model was built with logistic regression analysis and validated using external datasets.

Keratinocytes (KCs) expressing SLC2A1 and LDH1 were identified as a novel glycolysis-related subpopulation. Scissor⁺ cells and Scissor^- cells were defined as response and non-response phenotypes. In Scissor⁺ SLC2A1 KCs, ATP synthesis pathway was activated, especially, the glycolysis pathway being intriguing. Based on the glycolysis signature, keratinocyte differentiation was decomposed into a three-phase trajectory of normal, non-lesional, and lesional psoriatic cells. The area under the curve (AUC) and Brier score (BS) were used to estimate the performance of the glycolysis signature in distinguishing response and non-response samples in GSE69967 (AUC =0.786, BS =17.7) and GSE85034 (AUC=0.849, BS=11.1). Furthermore, Decision Curve Analysis suggested that the glycolysis score was clinically practicable.

We demonstrated a novel glycolysis-related subpopulation of KCs, identified 12-glycolysis signature, and validated its promising predictive efficacy of treatment effectiveness.