Table 1.
Full Biomarker panel.
Fig 1.
Workflow for biomarker-based clustering and recursive feature addition to characterise cardiovascular phenotypes in HIV.
Table 2.
Sequential addition of biomarkers (25–55) used in Model 1 to the initial 24-marker model, ordered by functional relevance to biological pathways identified in initial clustering.
Table 3.
Baseline demographics of combined cohorts.
Table 4.
Comparison of Baseline and RFA Models: Key Biomarkers and CVP associations and regression outcomes by Cluster.
Fig 2.
Heatmap showing biomarker contribution to cluster formation in the initial model.
Fig 3.
Baseline 24-biomarker model: associations with CVP.
Table 5.
Biomarker selection criteria scores model 1.
Fig 4.
Heatmap showing biomarker contribution to cluster formation in RFA model 1.
Table 6.
Median ASCVD 10-year risk and fold-change associations with CVP across models and clusters.
Fig 5.
Unadjusted and adjusted odd ratios (95% CI) for CVP in all three RFA models, using cluster 1 as the reference group.
Table 7.
Biomarker selection criteria scores model 2.
Fig 6.
Heatmap showing biomarker contribution to cluster formation in model 2.
Table 8.
Model 3 (Greedy forward - backward) biomarker selection path and resampling stability.
Fig 7.
Correlation Heatmap showing biomarker contribution to cluster formation in model 3.
Fig 8.
Overlap of Biomarkers Across the Three Predictive Models.
Table 9.
Comparison of biomarker selection robustness, cluster stability, and bootstrap regression results across models.
Table 10.
Distribution of study cohorts across clusters for each clustering model.