Machine learning gene expression predicting model for ustekinumab response in patients with Crohn's disease

• Published in: Immunity, Inflammation and Disease Inflammation and Disease, 2021-12, Vol.9 (4), p.1529-1540
• Main Authors: He, Manrong, Li, Chao, Tang, Wanxin, Kang, Yingxi, Zuo, Yongdi, Wang, Yufang
• Format: Article
• Language: English
• Subjects: Chemokines; Crohn Disease - drug therapy; Crohn Disease - genetics; Crohn's disease; Cytochrome; Datasets; Gene Expression; Generalized linear models; Humans; Infections; Inflammatory bowel disease; LASSO regression; Machine Learning; machine learning model; Metabolism; Monoclonal antibodies; Original; Regression analysis; Salmonella; Small intestine; Software; ustekinumab; Ustekinumab - therapeutic use
• ISSN: 2050-4527; 2050-4527
• DOI: 10.1002/iid3.506

Background Recent studies reported the responses of ustekinumab (UST) for the treatment of Crohn's disease (CD) differ among patients, while the cause was unrevealed. The study aimed to develop a prediction model based on the gene transcription profiling of patients with CD in response to UST. Methods The GSE112366 dataset, which contains 86 CD and 26 normal samples, was downloaded for analysis. Differentially expressed genes (DEGs) were identified first. Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analyses were administered. Least absolute shrinkage and selection operator regression analysis was performed to build a model for UST response prediction. Results A total of 122 DEGs were identified. GO and KEGG analyses revealed that immune response pathways are significantly enriched in patients with CD. A multivariate logistic regression equation that comprises four genes (HSD3B1, MUC4, CF1, and CCL11) for UST response prediction was built. The area under the receiver operator characteristic curve for patients in training set and testing set were 0.746 and 0.734, respectively. Conclusions This study is the first to build a gene expression prediction model for UST response in patients with CD and provides valuable data sources for further studies. In the present study, we used bioinformatics analysis to obtain the gene expression profiles from the Gene Expression Omnibus to build a model for UST response prediction. As a result, the gene expression profiling was revealed, and a multivariate logistic regression equation that comprises four genes (HSD3B1, MUC4, CF1, and CCL11) for UST response prediction was successfully built and evaluated. This study is the first to build a machine learning gene expression prediction model for UST response in patients with CD and provides valuable data sources for further basic and clinical studies in the future.