Elucidating the role of LGALS3BP in coronary atherosclerosis: integrating bioinformatics and machine learning for advanced insights | Journal of Cardiothoracic Surgery

DEG identification and principal component analysis

We integrated GSE43292 and GSE9820 and conducted batch match evidence integration. PCA corroborated the successful demarcation of patients into risk-specific cohorts (Fig. 2a, b). Among the 2798 DEGs, some DEGs were found to be significantly different. In addition, Some genes cluster in the AS group and some in the control group. Treat: DPP3, ADCY4, VAV1, CNDP2, RIPK3, ITGAL, PLCB2, PSME1, TAP1, PARP12, STAT1, etc. Control: CALD1, MYLK, MYL9, WASF3, PDE5 A, SPARC, PGRMC1, PROS1, MAOB, CDC14B, etc. (Fig. 2c). Some of these DEGs were significantly up-regulated (C1QB and C1QA). However, some genes were significantly down-regulated (SPARC, PROS1, CTNNAL1, GNG11, MEIS1, FSTL1, SH3BGRL2, C1QB, JAM3) (Fig. 2d) (Table.S1).

Construction of the model

To construct a robust gene signature for AS, we employed LASSO and Cox regression analyses to optimize gene selection, as demonstrated in Fig. 3a and b. Subsequently, the SVM-RFE technique was utilized to develop a machine learning model, confirming the model’s high accuracy and reliability with an accuracy rate of 0.696 and an error rate of 0.304 (Fig. 3c and d). Further investigation using Random Forest analysis identified several key genes (Fig. 3e). These DEGs were then analyzed through a comprehensive approach using Lasso regression, SVM-RFE, and Random Forest algorithms. This integrated analytical strategy successfully pinpointed 23 critical hub genes, confirmed by consensus among the outputs from the three methodologies (Fig. 3f; Table S2). Based on these results, LGALS3BP was selected for in-depth analysis, underlining its potential significance in the pathophysiology of AS (Table 2).

DEG identification and visualization

We visualized these 23 hub genes in the AS group and the normal sample group respectively (Fig. 4). In the confirmation of 23 hub genes, we analyzed the ROC of these genes, showing that the accuracy of these genes is high. SF3B3 (AUC:0.665), KLRD1 (AUC:0.639), ASPA (AUC:0.650), LGALS3BP (AUC:0.638), RABEP1 (AUC:0.614), STYK1 (AUC:0.621), SDK2 (AUC:0.622), NDFIP1 (AUC:0.613), IMPA1 (AUC:0.609), PKD1L3 (AUC:0.622), OR2 T8 (AUC:0.608), DAOA (AUC:0.596), C10orf111 (AUC:0.612), WDR77 (AUC:0.612), UBE2L3 (AUC:0.613), KCNA10 (AUC:0.600), KLHL10 (AUC:0.504), KCTD10 (AUC:0.586), TAS2R38 (AUC:0.589), SPANXB1 (AUC:0.591), OR56B4 (AUC:0.569), LRRC19 (AUC:0.583), MT1H (AUC:0.545) (Fig. 5).

Validation of hub genes

GSE9820 was used for validation to boost our model’s confidence and prediction accuracy of these hub genes. What’s interesting is that these DEGs are showed significant differences in GSE9820 analysis (Fig. 6a). In the GSE9820 analysis of hub genes, we analyzed the ROC of these genes, showing that the accuracy of these genes is high. SF3B3 (AUC:0.831), KLRD1 (AUC:0.741), ASPA (AUC:0.774), LGALS3BP (AUC:0.750), STYK1 (AUC:0.752), SDK2 (AUC:0.755), NDFIP1 (AUC:0.794), PKD1L3 (AUC:0.714), WDR77 (AUC:0.748), KCTD10 (AUC:0.820), TAS2R38 (AUC:0.723. These results also confirmed the high reliability and accuracy of our model (Fig. 6b).

Differential expression analysis centered on LGALS3BP

LGALS3BP was selected as a key investigative gene to determine its unique contributions to AS. Utilizing differential expression analysis focused on this gene, we identified 10 DEGs linked to LGALS3BP, as illustrated in Fig. 7. These DEGs varied significantly in expression, with certain genes predominating in distinct expression clusters. The ‘high’ expression cluster included notable genes such as ZBP1, SIGIRR, AES, FLT3LG, CD7, LGALS3BP, IFITM1, IL12RB1, FYN, HKDC1. Conversely, the ‘low’ expression cluster comprised genes like CD9, CORO1 C, THAP10, TST, FADS1, SCCPDH, ADAM28, ATP6 V0 A1, CHKA (Fig. 7a-b). Additionally, we developed a correlation matrix to further explore the relationships between LGALS3BP and these DEGs, providing a detailed visualization of these associations (Fig. 7c) and catalogued in Supplementary Table S3. This analysis not only highlights the differential roles of LGALS3BP in AS but also maps its potential regulatory network, offering insights into its biological and pathological roles.

Enrichment analysis of DEGs of LGALS3BP

Figure 8 presents the differential expression analysis of LGALS3BP, a precursor to functional enrichment analyses aimed at elucidating its biological roles. GO enrichment identified 65 principal targets categorized under MF and BP. The MF category was predominantly associated with enzyme inhibitor activity, carbohydrate binding, immune receptor activity. BPs included response to response to mononuclear cell differentiation, activation of immune response, lymphocyte differentiation. CC was predominantly associated with external side of plasma membrane, secretory granule membrane, cytoplasmic vesicle lumen. Additionally, KEGG pathway analysis highlighted significant involvement of overexpressed genes in pathways such as Cytokine-cytokine receptor interaction (hsa04060), PI3 K-Akt signaling pathway (hsa04151), Osteoclast differentiation (hsa04380). These findings, illustrated in Fig. 7 and Table S4a-b, provide insights into the molecular mechanisms through which LGALS3BP influences immune and inflammatory responses, potentially impacting AS pathology.

GSEA of analysis

To elucidate the biological functions impacted by differential expression of LGALS3BP, we applied GSEA using an array of computational tools including limma for differential expression analysis, org.Hs.eg.db for gene annotation, clusterProfiler and enrichplot for visualization of enrichment results. GSEA facilitated the identification of significant functional alterations among the DEGs linked to LGALS3BP. In the analysis of BP cytoplasmic translation, BP t cell activation, BP t cell differentiation. Conversely, in the low expression group, functional enrichments were noted inCC secretory granule membrane, CC specific granule, CC tertiary granule (Fig. 9a). KEGG pathway analysis revealed that high LGALS3BP expression groups showed significant enrichment in pathways like jak stat signaling pathway, ribosome, t cell receptor signaling pathway. Low expression groups exhibited enrichment in pathways associated with leishmania infection, lysosome, ppar signaling pathway (Fig. 9b, Table S4).

GSVA of analysis

For the GSVA, we utilized tools such as reshape2 for data restructuring, ggpubr for publication-quality visualizations, along with limma, GSEABase, and GSVA packages for robust analytical assessments. This analysis aimed to pinpoint functional alterations in the LGALS3BP DEGs. In GO terms, the high expression group showed enrichment in processes such as MF cytoskeleton nuclear membrane anchor activity, BP negative regulation of steroid metabolic process, BP glomerular epithelium development, BP actin polymerization dependent cell motility, BP inner ear auditory receptor cell differentiation, BP male gamete generation, CC aggresome (Fig. 10a). KEGG analysis highlighted enrichment in pathways including basal transcription factors, glycosylphosphatidylinositol gpi anchor biosynthesis, hedgehog signaling pathway, progesterone mediated oocyte maturation, ribosome, biosynthesis of unsaturated fatty acids (Fig. 10b).

Immune landscape characterization

Figure 11 explore the immune landscape of AS, with a particular focus on LGALS3BP as a pivotal gene for investigating its role within the immune contexture of the disease. This analysis provides vital insights into patterns of immune infiltration, underscoring the immunological factors critical to the initiation and progression of AS. The analysis revealed significant disparities in immune cell infiltration linked to risk profiles associated with LGALS3BP expression. In the LGALS3BP-defined cohorts, marked differences were observed in the infiltration levels of T helper cells between the low and high-risk groups. These variances highlight a complex immune modulation in different risk strata. Conversely, aDCs, APC co inhibition, B cells, CCR, HLA, Mast cells, Neutrophils, NK cells, Parainflammation, pDCs, Treg, Type I IFN Reponse, Check−point did not exhibit significant differences in infiltration between the risk groups, indicating a consistent involvement across the spectrum (P>0.05) as detailed in Fig. 11a. This dichotomy in immune cell behavior underscores the nuanced role of LGALS3BP in modulating the immune environment in AS. B cells naive, Plasma cells, T cells CD4 memory resting, T cells CD4 memory activated, Macrophages M2, Dendritic cells activated, Neutrophils were highly expressed in the treat group. While, Dendritic cells resting, T cells follicular helper, T cells regulatory (Tregs), T cells gamma delta, NK cells resting, B cells memory did not exhibit significant differences in infiltration between the risk groups (Fig. 11b). In addition, we also constructed an immune infiltration correlation rectangle plot and heatmap (Fig. 11c, d). Through PCA analysis, immune-based patient categorization was again successfully executed (Fig. 11e). A Lollipop was created to display the expression patterns of Correlation Coefficient. T cells CD8, NK cells activated, T cells follicular helper, T cells CD4 naive, T cells gamma delta, T cells regulatory (Tregs), T cells CD4 memory resting (Fig. 11f). Dendritic cells activated, NK cells activated, T cells CD4 memory resting, T cells CD4 naive, T cells CD8, T cells follicular helper, T cells gamma delta, T cells regulatory (Tregs) were shown to be positively associated with LGALS3BP, While, Mast cells resting, Macrophages M0, Eosinophils were shown to be negatively associated with LGALS3BP (Fig. 12) (Table.S5).

Identification of common RNAs and construction of miRNAs-LncRNAs shared genes network

Figure 13 uses LGALS3BP as a hub gene to investigate its expression dynamics within AS-associated miRNAs and lncRNAs, aiming to delineate its regulatory network. Three databases were searched for 6 miRNAs and 24 lncRNAs linked with AS (Table.S7a-b). The network of miRNAs-lncRNAs-genes was constructed by taking the intersection of them and shared genes (obtained by Lasso regression and SVM-RFE). Finally, the miRNAs-genes network included 23 lncRNAs (RP11-10 J21.4, RP11-830 F9.6, LINC00917, RP5-894D12.5, RP1-34P24.3, AC097468.4, CH507-216 K13.2, etc), 3 miRNAs (hsa-miR-590-3p, hsa-miR-671-5p, hsa-miR-125a-3p) (Fig. 13) (Table S6).