Machine learning predicts relapse after anti-TNF withdrawal in ulcerative colitis using rectal biopsy transcriptomics
- 1Oslo University Hospital
Abstract
BackgroundThere are currently no available biomarkers that predict the clinical outcome of ulcerative colitis (UC) patients after anti-TNF withdrawal. Here we developed a semi-supervised machine learning framework on RNA-seq data derived from rectal biopsies to predict relapse following anti-TNF discontinuation.
MethodsWe included UC patients (n=136) enrolled in the prospective, randomized controlled BIOSTOP trial. Patients were assigned to a withdrawal group (stopped treatment immediately) or a maintenance group (stopped treatment after 2 years). A machine learning framework was trained on gene expression data from mucosal biopsies collected at baseline from the withdrawal group and validated in the maintenance group using baseline samples (test cohort) and 2-year follow-up samples (interference cohort).
ResultsA semi-supervised machine learning model, combining a shared UMAP manifold representation with LabelSpreading on a targeted 38-gene panel achieved an AUC of 0.794, a sensitivity of 90.3%, and a specificity of 73.1% on the test cohort, and an AUC of 0.711, a sensitivity of 82.9%, and a specificity of 42.3% on the inference cohort. Applying Shapley Additive exPlanations (SHAP) we found that IL1B, TYROBP, REG1B, S100A8, and TNF showed the highest predictive importance for relapse risk, whereas MUC2, HMGCS2, and CDKN1A displayed the highest predictive importance for sustained remission.
ConclusionThe high sensitivity makes the model a promising screening tool to prevent withdrawal-related relapse, though its moderate specificity underscores the need for multi-center validation in larger, more diverse cohorts before clinical deployment.
Related
- A mucosal barrier injury related six gene transcriptomic signature associated with ulcerative colitis using machine learning and public transcriptomic data
- Degree bias predicts, outperforms and improves machine learning phage-host interaction predictions
- Integrative Transcriptomics and Machine Learning Identify Temporal Stress Signatures and Astrocytic Neurocan Associated With Hypothalamic Injury.
- Single-cell transcriptomics and machine learning characterize LTA4H-linked lipid metabolism and inflammatory signatures in the knee osteoarthritis IPFP/synovium compartment
- Identification of psoriasis-associated hub genes by integrated transcriptomics and machine learning
BibTeX
@misc{bouhddi2026machine,
title = {{Machine learning predicts relapse after anti-TNF withdrawal in ulcerative colitis using rectal biopsy transcriptomics}},
author = {G. Bouhddi and E. Rapp and P. Artiemjew and I. P. Berset and F. Lerang and D. A. L. Hoff and E. S. Bakkevold and K. E. A. Lundin and F. L. Jahnsen and D. Domanska},
howpublished = {medRxiv},
year = {2026},
month = oct,
doi = {10.64898/2026.10.05.26364793},
url = {https://www.medrxiv.org/content/10.64898/2026.10.05.26364793v1},
note = {CC BY-NC},
}
RIS
TY - UNPB
TI - Machine learning predicts relapse after anti-TNF withdrawal in ulcerative colitis using rectal biopsy transcriptomics
AU - G. Bouhddi
AU - E. Rapp
AU - P. Artiemjew
AU - I. P. Berset
AU - F. Lerang
AU - D. A. L. Hoff
AU - E. S. Bakkevold
AU - K. E. A. Lundin
AU - F. L. Jahnsen
AU - D. Domanska
PY - 2026
DA - 2026/10/05
PB - medRxiv
DO - 10.64898/2026.10.05.26364793
UR - https://www.medrxiv.org/content/10.64898/2026.10.05.26364793v1
AB - BackgroundThere are currently no available biomarkers that predict the clinical outcome of ulcerative colitis (UC) patients after anti-TNF withdrawal. Here we developed a semi-supervised machine learning framework on RNA-seq data derived from rectal biopsies to predict relapse following anti-TNF discontinuation. MethodsWe included UC patients (n=136) enrolled in the prospective, randomized controlled BIOSTOP trial. Patients were assigned to a withdrawal group (stopped treatment immediately) or a maintenance group (stopped treatment after 2 years). A machine learning framework was trained on gene expression data from mucosal biopsies collected at baseline from the withdrawal group and validated in the maintenance group using baseline samples (test cohort) and 2-year follow-up samples (interference cohort). ResultsA semi-supervised machine learning model, combining a shared UMAP manifold representation with LabelSpreading on a targeted 38-gene panel achieved an AUC of 0.794, a sensitivity of 90.3%, and a specificity of 73.1% on the test cohort, and an AUC of 0.711, a sensitivity of 82.9%, and a specificity of 42.3% on the inference cohort. Applying Shapley Additive exPlanations (SHAP) we found that IL1B, TYROBP, REG1B, S100A8, and TNF showed the highest predictive importance for relapse risk, whereas MUC2, HMGCS2, and CDKN1A displayed the highest predictive importance for sustained remission. ConclusionThe high sensitivity makes the model a promising screening tool to prevent withdrawal-related relapse, though its moderate specificity underscores the need for multi-center validation in larger, more diverse cohorts before clinical deployment.
ER -