Identification of Ferroptosis-Related Genes in Type 2 Diabetes Mellitus
Based on the Machine Learning
Abstract
Background: Type 2 diabetes mellitus (T2DM), which has a high incidence
and several harmful consequences, poses a severe danger to human health.
More research is being done on ferroptosis’ function in T2DM. This study
uses a bioinformatics technique to look for new diagnostic T2DM
biomarkers associated with ferroptosis. Methods: In order to identify
ferroptosis-related genes (DEGs) that are differently expressed between
T2DM patients and healthy individuals, we first obtained T2DM sequencing
data and ferroptosis-related genes (FRGs) from the Gene Expression
Omnibus (GEO) database and FerrDb database. Then, drug-gene interaction
networks and ceRNA networks linked to the marker genes were built after
marker genes were filtered by two machine learning algorithms (LASSO and
SVM-RFE algorithms). Finally, to confirm the expression of marker genes,
the GSE76895 dataset was utilized. The protein expression of some marker
genes between T2DM and non-diabetic tissues was also examined by Western
Blotting, Immunohistochemistry (IHC) and Immunofluorescence (IF),
respectively. Results: We obtained 58 DEGs associated with ferroptosis.
GO and KEGG enrichment analysis showed that these DGEs were
significantly enriched in hypoxia and ferroptosis. Subsequently, eight
marker genes (SCD, CD44, HIF1A, BCAT2, MTF1, HILPDA, NR1D2 and MYCN)
were screened by LASSO and SVM- RFE machine learning algorithms, and a
model was constructed based on these eight genes. These newly discovered
marker genes may be linked to alterations in the immune microenvironment
in T2DM patients. In addition, based on these 8 genes, we obtained 48
drugs and a complex ceRNA network map. Finally, Western Blotting, IHC
and IF results of clinical samples further confirmed the results of
public databases. Conclusions: The diagnosis and etiology of T2DM can be
greatly aided by eight ferroptosis-related genes, opening up novel
therapeutic avenues.