Identification of clinical subgroups in anti-SRP positive immune-mediated necrotizing myopathy patients using cluster analysis

Figures

• 

  Figure 1.. K-means clustering analysis results. (a) The x-axis represents the number of clusters, while the y-axis represents the silhouette width. Peaks (k = 3) in the plot suggest the optimal number of clusters, with an average silhouette width value of 0.194. (b) This scatter plot visualizes the clustering of data points based on the k-means algorithm. Each point represents an enrolled patient, plotted with respect to variables X and Y. The different colors (red, green, blue) correspond to different clusters, labeled as 1, 2, and 3, respectively.

• 

  Figure 2.. Kaplan–Meier analysis of relapse-free survival probability. Cluster 1 is represented by the brown line, Cluster 2 by the green line, and Cluster 3 by the blue line. The plot indicates the number of patients at risk at different time points for each cluster. No statistical differences were observed between the three clusters overall (p = 0.100). However, a significant difference was found between Cluster 1 patients and non-Cluster 1 patients (p# = 0.035).

• 

  Figure 3.. Heatmap of cluster analysis. This heatmap illustrates the clustering results of various variables across three clusters. The rows represent different variables such as age, male, ILD, CardInj, High ANA (antinuclear antibody, ⩾1:1000), DRR (AST/ALT), anti-SRP antibodies, UA, IgA, IgG, MRC ⩽3, CK, LDH, IgM, and Rash. The columns represent the three identified clusters (Cluster 1, Cluster 2, and Cluster 3). The color gradient indicates the normalized values of the variables, with red representing higher values, blue representing lower values, and intermediate colors representing values in between. ALT, alanine aminotransferase; ANA, antinuclear antibodies; anti-SRP, anti-signal recognition particle autoantibodies; AST, aspartate aminotransferase; CardInj, cardiac injury; CK, creatine kinase; DRR, De Ritis Ratio; IgA, immunoglobulin A; IgG, immunoglobulin G; IgM, immunoglobulin M; ILD, interstitial lung disease; LDH, lactate dehydrogenase; MRC, modified Rankin Scale; UA, uric acid.

• 

  Figure 4.. Decision tree model for key feature distinction. A decision tree model was trained to distinguish key features from clusters with an accuracy of 0.784 (95% CI: 0.684, 0.865). The root node starts with ANA, splitting into two branches based on ANA levels (<1:1000 and ⩾1:1000). The left branch, for ANA levels below 1:1000, further splits based on age, differentiating between individuals aged 55 and older versus those younger than 55. Conversely, the right branch, for ANA levels 1:1000 and above, divides by gender into male and female subsets. Among males, an additional split is introduced, evaluating the DRR (or AST/ALT), with a cutoff of 0.94 to further categorize individuals. The terminal nodes of this decision tree, color-coded in red, blue, and green, represent distinct clusters, each with its associated probabilities and cluster proportions. ALT, alanine aminotransferase; ANA, antinuclear autoantibodies; AST, aspartate aminotransferase; DRR, De Ritis Ratio.