Researchers from Greece developed a machine learning model able to identify a microRNA (miRNA) expression profile that could aid in gastrointestinal stromal tumor (GIST) risk assessment and published their findings in Cancer Biomarkers.
They trained and validated 7 classification models based on differentially expressed miRNAs, and all models showed high accuracy for GIST risk prediction. The analysis of gastric GISTs (n=20) and healthy gastric (n=20) biopsies revealed 56 differentially expressed miRNAs.
Three miRNAs (miR-196a-5p, miR-148a-3p, and miR-125a-5p) were upregulated in tumor samples, while the others were downregulated (including miR-218-5p, miR-126-3p, miR-222-3p, and hsa-miR-146a-5p). Most of these miRNAs had been linked to GIST in previous studies.
Read more about GIST testing
“These miRNAs directly target fundamental genes in GIST pathogenesis like KIT/AKT, [platelet-derived growth factor receptor A] pathways, and have also been found as crucial carcinogenesis mediators in other gastrointestinal cancers such as gastric cancer,” the authors said.
The researchers used the differentially expressed miRNAs to construct a machine learning classifier. miR-218-5p, a known tumor suppressor, was identified as the best predictor for GIST development in the study cohort. According to other studies, miR-218 increases cell sensitivity to imatinib mesylate.
miR-146a-5p, miR-222-3p, and miR-126-3p were also identified as top predictors in all the tested models. The role of these miRNAs in GISTs is not fully elucidated.
“Even though we performed comprehensive machine learning and bioinformatics analyses using the miRNA expression profile of GIST and confirmed the classification accuracy by cross-validation, there are some limitations in our study,” the authors said. Hence, these results seek further validation.
Stefanou IK, Dovrolis N, Gazouli M, Theodorou D, Zografos GK, Toutouzas KG. miRNAs expression pattern and machine learning models elucidate risk for gastric GIST. Cancer Biomarkers. 2022;33(2):237-247. doi:10.3233/CBM-210173