转录组学数据的共表达网络与机器学习分析确定了特应性皮炎皮损和非皮损部位不同的基因特征及通路

Co-Expression Network and Machine Learning Analysis of Transcriptomics Data Identifies Distinct Gene Signatures and Pathways in Lesional and Non-Lesional Atopic Dermatitis.

作者信息

Dessie Eskezeia Y, Ding Lili, Satish Latha, Mersha Tesfaye B

机构信息

Division of Asthma Research, Cincinnati Children's Hospital Medical Center, Department of Pediatrics, University of Cincinnati College of Medicine, 3333 Burnet Avenue, Cincinnati, OH 45229-3039, USA.

Division of Biostatistics and Epidemiology, Cincinnati Children's Hospital Medical Center, Department of Pediatrics, University of Cincinnati College of Medicine, 3333 Burnet Avenue, Cincinnati, OH 45229-3039, USA.

出版信息

J Pers Med. 2024 Sep 10;14(9):960. doi: 10.3390/jpm14090960.

DOI:10.3390/jpm14090960

PMID:39338214

原文链接:https://pmc.ncbi.nlm.nih.gov/articles/PMC11433539/

Abstract

BACKGROUND

Atopic dermatitis (AD) is a common inflammatory skin condition with complex origins. Current treatments often yield suboptimal results due to an incomplete understanding of its underlying mechanisms. This study aimed to identify pathway and gene signatures that distinguish between lesional AD, non-lesional AD, and healthy skin.

METHOD

We conducted differential gene expression and co-expression network analyses to identify differentially co-expressed genes (DCEGs) in lesional AD vs. healthy skin, lesional vs. non-lesional AD, and non-lesional AD vs. healthy skin. Modules associated with lesional and non-lesional AD were identified based on the correlation coefficients between module eigengenes and clinical phenotypes (|R| ≥ 0.5, -value < 0.05). Subsequently, we employed Ingenuity Pathway Analysis (IPA) on the identified DCEGs, followed by machine learning (ML) analysis within the pathway expression framework. The ML analysis of pathway expressions, selected by IPA and derived from gene expression data, identified relevant pathway signatures, which were validated using an independent dataset and correlated with AD severity measures (EASI and SCORAD).

RESULTS

We identified 975, 441, and 40 DCEGs in lesional vs. healthy skin, lesional vs. non-lesional, and non-lesional vs. healthy skin, respectively. IPA and ML analyses revealed 25 relevant pathway signatures, including wound healing, glucocorticoid receptor signaling, and S100 gene family signaling pathways. Validation confirmed the significance of 10 pathway signatures, which were correlated with the AD severity measures. DCEGs such as MMP12 and S100A8 demonstrated high diagnostic efficacy (AUC > 0.70) in both the discovery and validation datasets.

CONCLUSIONS

Differential gene expression, co-expression networks and ML analyses of pathway expression have unveiled relevant pathways and gene signatures that distinguish between lesional, non-lesional, and healthy skin, providing valuable insights into AD pathogenesis.

摘要

背景

特应性皮炎（AD）是一种常见的炎症性皮肤病，病因复杂。由于对其潜在机制的理解不完整，目前的治疗效果往往不理想。本研究旨在识别区分AD皮损、非皮损和健康皮肤的通路和基因特征。

方法

我们进行了差异基因表达和共表达网络分析，以识别AD皮损与健康皮肤、皮损与非皮损AD、非皮损AD与健康皮肤之间的差异共表达基因（DCEG）。基于模块特征基因与临床表型之间的相关系数（|R|≥0.5，P值<0.05），识别与皮损和非皮损AD相关的模块。随后，我们对鉴定出的DCEG进行了 Ingenuity 通路分析（IPA），然后在通路表达框架内进行机器学习（ML）分析。通过IPA选择并从基因表达数据中得出的通路表达的ML分析确定了相关的通路特征，并使用独立数据集进行了验证，并与AD严重程度指标（EASI和SCORAD）相关联。

结果

我们分别在皮损与健康皮肤、皮损与非皮损、非皮损与健康皮肤中鉴定出975、441和40个DCEG。IPA和ML分析揭示了25个相关的通路特征，包括伤口愈合、糖皮质激素受体信号传导和S100基因家族信号通路。验证证实了10个通路特征的重要性，这些特征与AD严重程度指标相关。MMP12和S100A8等DCEG在发现和验证数据集中均显示出较高的诊断效能（AUC>0.70）。