Analyzing immune cell infiltrates in skeletal muscle of infantile-onset Pompe disease using bioinformatics and machine learning

Abstract

Pompe disease, a severe lysosomal storage disorder, is marked by heart problems, muscle weakness, and respiratory difficulties. This study aimed to identify novel markers for infantile-onset Pompe disease by analyzing key genes and immune cells infiltrating affected skeletal muscles, building on existing research linking its progression to the immune cell infiltration. The datasets GSE38680 and GSE159062 were downloaded and differential expression genes (DEGs) were identified. Potential markers were screened using support vector machine recursive feature elimination (SVM-RFE) analysis and least absolute shrinkage and selection operator (LASSO) regression models. To analyze DEGs and immune processes, 22 immune cell types in affected tissues were examined using CIBERSORT. Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analysis revealed significant enrichment in the calcium signaling pathway, phosphatidylinositol signaling system, ubiquitin-mediated proteolysis and JAK-STAT signaling pathway. Machine learning identified GPNMB, CALML6 and TRIM7 as key genes. Immune cell infiltration analysis showed that the expression levels of the above three genes were closely related to immune cells in our target population under study, including T cells regulatory (Tregs), monocytes, macrophages M0, resting and activated dendritic cells, naive and memory B cells. Overall, our study results may provide new clues for exploring markers related to Pompe disease by highlighting key genes and their relationship with immune infiltration, offering insights into the disease's development. However, previous studies have shown limited evidence of significant immune cell infiltration in muscle biopsies from Pompe patients, and this lack of direct evidence necessitates further investigation and experimental validation in laboratory settings to confirm these associations.

Keywords: GEO datasets; Immune cells infiltration; Machine learning; Pompe disease; Signatures.

Fig. 1

Volcano plot (a) and heat map (b) of DEGs between Pompe patients and control samples. (a) Dashed lines define significance thresholds for differential expression. The x-axis, Log2FC (Log2 Fold Change), shows gene expression changes, with positives indicating upregulation and negatives indicating downregulation. The y-axis, Log10FDR (False Discovery Rate), plots the False Discovery Rate’s logarithm, where lower values denote higher significance. (b) The x-axis represents each sample, combining two GEO datasets followed by cluster analysis. The gene expression levels are displayed as normalized values on a scale, where negative values indicate downregulation relative to a reference point, and positive values indicate upregulation.

Fig. 2

GO analysis (a) and KEGG analysis (b) of 38 DEGs via the ClusterProfile. Figure (a) displays the results of the Gene Ontology (GO) analysis for differentially expressed genes. The left side of the figure categorizes the results into Biological Process, Molecular Function, and Cellular Component. The size of each circle represents the number of genes enriched in each category, and the color gradient from blue to pink indicates progressively smaller p-values. Figure (b) displays the results of the KEGG analysis. The left side shows the names of the KEGG metabolic or signal transduction pathways, and the circle sizes have the same meaning as in the GO analysis.

Fig. 3

Selection of candidate signature for Pompe disease: (a) tuning feature screening in the LASSO model; (b) the accuracy and error rate of curve changes after 10x cross validation via the SVM-RFE arithmetic; (c) Venn graph displaying 3 signature genes shared by LASSO and SVM-RFE.

Fig. 4

The expression of GPNMB, CALML6, and TRIM7 in Pompe disease: (a) GPNMB expression was distinctly upregulated in Pompe samples; (b, c) The expressions of CALML6 and TRIM7 in Pompe samples were significantly downregulated.

Fig. 5

(a) The percentage of the 22 immunocyte populations identified via the CIBERSORT arithmetic. The X-axis features pink and blue triangles representing controls and Pompe disease patients, respectively. (b) Displays correlations between different infiltrative immune cell populations in skeletal muscle samples from Pompe disease patients and controls. The strongest positive correlation is between monocytes and macrophages M0, and the strongest negative correlation is between T cells CD4 memory resting and T cells follicular helper. (c) The diversities in the architecture of immunocytes between controls and Pompe patients.

Fig. 6

Correlation between GPNMB (a), CALML6 (b), TRIM7 (c) expression levels and infiltrating immune cells in muscle tissue of infantile-onset Pompe patients. The left side of the figure displays the names of immune cells. The larger the circle in the figure, the greater the correlation with the immune cells. The color transition from red to green in the circles indicates increasingly smaller p-values. Specific p-values are displayed on the right side of the figures.