Additional file 1 of Exploring synthetic controls in rare diseases with a proof of concept in spinal cord injury
Bibliographic record
Abstract
Additional file 1: Sections 1–2. Section 1 — Supplementary Methods. Section 2 — Supplementary Results. Figures S1-S6. Figure S1 — Schematic overview of the deep learning architectures. Figure S2 — Consort diagram for EMSCI cohort. Figure S3 — Consort diagram for Sygen cohort. Figure S4 — RMSEbl.NLI as function of time of initial assessment. Figure S5 — RMSEbl.NLI as function of time of initial assessment for EMSCI cohort subset according to NISCI inclusion criteria. Figure S6 — Importance ranking of interpretability SHAP scores. Tables S1-S15. Table S1 — Hyperparameters for tree-based models. Table S2 — Hyperparameters for deep learning models. Table S3 — Characteristics of EMSCI and Sygen cohorts used for machine learning benchmark in comparison with subsets excluded. Table S4 — Number of instances in EMSCI with missing age. Table S5 — Number of instances in EMSCI within each AIS grade with imputed VAC. Table S6 — Number of instances in EMSCI within each AIS grade with imputed DAP. Table S7 — Results of the model benchmark. Table S8 — Performance on the EMSCI dataset stratified by AIS grade. Table S9 — Performance on the Sygen dataset stratified by AIS grade. Table S10 — Median of mean residuals below the NLI. Table S11 — Median of mean residual below NLI on the EMSCI dataset stratified by AIS grade. Table S12 — Median of mean residual below NLI on the Sygen dataset stratified by AIS grade. Table S13 — Benchmark on EMSCI cohort with NISCI inclusion criteria. Table S14 — Benchmark of CNN multi-modal trained on the same data as before and all combinations of time points. Table S15 — Distribution of group-level differences in mean LEMSimpr.
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.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.004 | 0.067 |
| Meta-epidemiology (narrow) | 0.001 | 0.001 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.002 | 0.002 |
| Science and technology studies | 0.001 | 0.000 |
| Scholarly communication | 0.002 | 0.001 |
| Open science | 0.002 | 0.001 |
| Research integrity | 0.002 | 0.001 |
| Insufficient payload (model declined to judge) | 0.746 | 0.098 |
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 itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".