Figures (5)  Tables (0)

    • Figure 1. 

      Conceptual framework of a latent disease axis for SLE.

    • Figure 2. 

      Disease-axis projection and paired treatment shifts. (a) Distribution of disease-axis projection scores for healthy controls (HC), untreated SLE patients (SLE), and post-treatment SLE samples (SLE_post). Higher scores indicate a more SLE-like transcriptional state. (b) Paired shifts along the disease axis for individual patients, comparing pre-treatment (SLE) and post-treatment (SLE_post) samples. Each line represents one patient. (c) Individual-level treatment trajectories, highlighting heterogeneous responses ranging from marked improvement to minimal change.

    • Figure 3. 

      Representation analysis revealing treatment-associated trajectories in the latent space. (a) PCA of normalized gene expression across HC, SLE (pre-treatment), and SLE_post (post-treatment). Ellipses denote 95% confidence regions. (b) Latent manifold embedding showing individual treatment vectors $ {\delta }_{i}={z}_{i,\text{SLET}}-{z}_{i,\text{SLE}} $. Vectors are colored by their projection onto the disease axis v (blue: $ \text{Δ}{s}_{i}> 0 $, red: $ \text{Δ}{s}_{i}\leq 0 $).

    • Figure 4. 

      Representative trajectory-defined modules identified by trivariate functional clustering across healthy controls (HC), untreated SLE, and SLE_post (post-treatment SLE).

    • Figure 5. 

      GO biological process enrichment of PRM, FRM, PM, and SM reveals functional layers of the disease continuum.