BioBERT-Derived Clinical TextRepresentations and Sparse PrincipalComponents of Multi-Omics Data for BreastCancer Survival Prediction: ALeakage-Controlled Cross-ValidatedBenchmark
- 1University of Central Florida
Abstract
BackgroundReported performance for multi-modal cancer prognostic models is frequently obtained from a single train-test partition and without comparison against a simple clinical baseline. We evaluated an integrated framework combining language-model representations of structured clinical records with high-dimensional multi-omics data under a leakage-controlled protocol.
MethodsClinical variables from 589 TCGA-BRCA patients were converted to natural language sentences and encoded with BioBERT; DNA methylation, RNA, miRNA and protein data were reduced by Sparse principal component analysis. Clinical-only, multi-omics-only and integrated feature sets were evaluated with Cox proportional hazards, XGBoost and random survival forest (RSF) using five repetitions of stratified five-fold cross-validation with pooled out-of-fold predictions. Models were compared against a marginal Kaplan-Meier null and a conventional age-and-stage Cox model, with differences assessed by paired bootstrap.
ResultsAmong 589 patients with 88 deaths (14.9%; median follow-up 33.7 months), the integrated RSF model performed best (Harrells C 0.685, 95% CI 0.620-0.754; Unos C 0.748), exceeding the conventional clinical Cox model (0.637, 0.559-0.707). The ordering integrated > clinical-only > multi-omics-only held in all nine feature-set by algorithm comparisons and at all three component settings tested. The integrated model significantly outperformed every multi-omics-only model (P = 0.008-0.044) but not clinical-only models ({Delta}C = 0.037, P = 0.128). Brier scores approached a null reference at 2 and 3 years and improved modestly at 5 years.
ConclusionIntegration produced the highest discrimination and exceeded a conventional clinical model, but the gain over clinical features within the same pipeline was not statistically resolvable at this event count. Reporting simple baselines, null-model calibration and cross-validated estimates should be standard in multi-modal prognostic modelling.
Related
- Baseline machine learning prediction of 2-year remission from anxiety, depression, and eating disorders among college students after population-based guided self-help: A secondary analysis of a randomized controlled trial
- Contralateral-controlled thigh muscle biomarkers on CT angiography for prediction of 1-year restenosis after endovascular therapy for unilateral peripheral artery disease
- Clinical Effectiveness of AI-Based Cardiac Arrest Prediction System for General Ward Patients (VITAL-MD): Protocol for a Multicenter Pragmatic Stepped Wedge Cluster Randomized Controlled Trial.
- Personalized assistive EMG-controlled virtual reality interface combined with cerebellar-iTBS for post-stroke arm recovery (ERICA): study protocol for a three-arm randomized controlled trial
- Acupuncture for older adults with dementia in residential care homes: an assessor-blinded, randomised controlled trial
BibTeX
@misc{bhuiyan2026biobertderived,
title = {{BioBERT-Derived Clinical TextRepresentations and Sparse PrincipalComponents of Multi-Omics Data for BreastCancer Survival Prediction: ALeakage-Controlled Cross-ValidatedBenchmark}},
author = {M. M. H. Bhuiyan},
howpublished = {medRxiv},
year = {2026},
month = aug,
doi = {10.64898/2026.08.31.26361875},
url = {https://www.medrxiv.org/content/10.64898/2026.08.31.26361875v1},
note = {CC BY-NC-ND},
}
RIS
TY - UNPB
TI - BioBERT-Derived Clinical TextRepresentations and Sparse PrincipalComponents of Multi-Omics Data for BreastCancer Survival Prediction: ALeakage-Controlled Cross-ValidatedBenchmark
AU - M. M. H. Bhuiyan
PY - 2026
DA - 2026/08/31
PB - medRxiv
DO - 10.64898/2026.08.31.26361875
UR - https://www.medrxiv.org/content/10.64898/2026.08.31.26361875v1
AB - BackgroundReported performance for multi-modal cancer prognostic models is frequently obtained from a single train-test partition and without comparison against a simple clinical baseline. We evaluated an integrated framework combining language-model representations of structured clinical records with high-dimensional multi-omics data under a leakage-controlled protocol. MethodsClinical variables from 589 TCGA-BRCA patients were converted to natural language sentences and encoded with BioBERT; DNA methylation, RNA, miRNA and protein data were reduced by Sparse principal component analysis. Clinical-only, multi-omics-only and integrated feature sets were evaluated with Cox proportional hazards, XGBoost and random survival forest (RSF) using five repetitions of stratified five-fold cross-validation with pooled out-of-fold predictions. Models were compared against a marginal Kaplan-Meier null and a conventional age-and-stage Cox model, with differences assessed by paired bootstrap. ResultsAmong 589 patients with 88 deaths (14.9%; median follow-up 33.7 months), the integrated RSF model performed best (Harrells C 0.685, 95% CI 0.620-0.754; Unos C 0.748), exceeding the conventional clinical Cox model (0.637, 0.559-0.707). The ordering integrated > clinical-only > multi-omics-only held in all nine feature-set by algorithm comparisons and at all three component settings tested. The integrated model significantly outperformed every multi-omics-only model (P = 0.008-0.044) but not clinical-only models ({Delta}C = 0.037, P = 0.128). Brier scores approached a null reference at 2 and 3 years and improved modestly at 5 years. ConclusionIntegration produced the highest discrimination and exceeded a conventional clinical model, but the gain over clinical features within the same pipeline was not statistically resolvable at this event count. Reporting simple baselines, null-model calibration and cross-validated estimates should be standard in multi-modal prognostic modelling.
ER -