IDENTIFYING HOMOGENOUS ENDOPHENOTYPES IN CHILDHOOD-ONSET SLE WITH DATA-DRIVEN METHODS
Why this work is in the frame
A frame that forgets how it found something cannot be audited. These are the routes that admitted this work.
Bibliographic record
Abstract
O072 / #800 Topic: AS18 - Pediatric SLE Late-Breaking Abstract ABSTRACT CONCURRENT SESSION 12: PEDIATRIC SLE – ADVANCES IN DISEASE OUTCOMES AND MENTAL HEALTH 24-05-2025 10:40 AM - 11:40 AM Background/Purpose Childhood-onset Systemic Lupus Erythematosus (cSLE) is a clinically heterogeneous autoimmune disease. We hypothesized that data-driven methods would identify clinically homogeneous patient subgroups that may represent cSLE endotypes with distinct genetics. Methods We included patients diagnosed with cSLE between January 1992-October 2023. All patients met 2019 ACR-EULAR classification criteria and were genotyped on Illumina multiethnic arrays. Ungenotyped single nucleotide polymorphisms were imputed with TopMed as a referent. We extracted SLE manifestations, date of each manifestation onset and demographics from dedicated Lupus databases. Ancestry was genetically inferred using principal components and ADMIXTURE with 1000 Genomes as a referent. We used time from SLE diagnosis to each manifestation to identify patient clusters using similarity network fusion (SNF), a data-driven method. We used Kaplan-Meier analyses and Cox proportional-hazard models to compare clusters. We tested cluster differences in demographic and manifestation prevalences using χ 2 or Fisher’s exact test, and time to each SLE manifestation onset with log-rank tests. Our clustering was validated with simulation-based sensitivity analysis (1000 iteration of simulated SNF). Each iteration randomly subsampled 70% of our cohort, performed SNF and tested cluster differences in demographics, manifestation prevalence and each SLE manifestation onset. Genetic studies tested 162 SLE associated genes from 3 transancestral SLE genome wide association studies and 33 monogenic SLE genes with cluster membership using sequence kernel association tests (SKAT). SKAT was weighted by minor allele frequency and adjusted for sex, ancestry and age of diagnosis. The threshold for significance was adjusted for multiple comparison with the Bonferroni correction ( P < 2.6 x10^-4; 0.05 / 195). Results Our cohort included 442 cSLE patients. 83% were female and the median age of SLE diagnosis was 13.6 years (Q1-Q3: 12.0-15.8). The majority of patients were of European (27%) and East Asian (26%) ancestry, followed by South Asian (18%), Admixed (17%) and African (12%) ancestry. SNF identified 2 clusters. Patients in cluster 1 (n = 205) were predominantly of European ancestry (42%), while cluster 2 (n = 237) was mainly composed of patients of East Asian (30%) and South Asian (22%) ancestry ( P = 3x10^-9). Patients in cluster 2 had higher prevalence of class III/IV lupus nephritis, fever, oral ulcers, hypocomplementemia, anemia, leukopenia, anti-cardiolipin and anti-Smith antibodies compared to patients in cluster 1 ( P < 1x10^-7; Figure 1). Moreover, patients in cluster 2 had an earlier onset of developing the same 9 SLE manifestations as the risk of developing each manifestation at any time was higher in cluster 2 compared to cluster 1 (HR > 1.4; P < 4x10^-3). Simulation-based sensitivity analysis demonstrated that the same 9 SLE manifestation consistently drove clustering (>900/1000 times) and 95% of patients consistently clustered together over 1000 simulations. None of the 195 SLE genes were associated with cluster membership. Figure 1: Clinical and Laboratory SLE Manifestation With Different Prevalences Between Patient Clusters . The number in each cell represents the prevalence of an SLE manifestation within cluster 1 and 2, respectively. “LN” stands for lupus nephritis. Statistics performed with a Fisher’s Exact Test, Bonferroni corrected P < 0.002. Conclusions In a large multiethnic cSLE cohort, data-driven methods identified 2 robust cSLE patient clusters. The cluster with more severe disease and younger onset had a greater proportion of patients of East Asian and South Asian ancestry compared to the cluster with milder disease. Simulation-based sensitivity analysis demonstrated that 95% of patients consistently clustered together and 9 SLE manifestation primarily determined these clusters. Future work elucidating the role of genetics in our clustering is needed.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
Full frame distilled prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.
Codex and Gemma teacher scores by category
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.001 | 0.000 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.000 | 0.001 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.000 | 0.000 |
| Open science | 0.001 | 0.000 |
| Research integrity | 0.000 | 0.000 |
| Insufficient payload (model declined to judge) | 0.000 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from it