Self-supervised plasma proteomic representations for prospective disease prediction across varying protein availability
- 1University of Pennsylvania
Abstract
Large-scale plasma proteomics offers opportunities to characterize disease susceptibility and improve prospective risk prediction, but transferring proteomic predictors across datasets remains challenging because measured protein sets differ across cohorts, study phases and assay configurations. Here we developed a self-supervised protein-token Transformer that maps the proteins observed in each sample to a fixed-dimensional participant representation, allowing unavailable proteins to be omitted rather than imputed. Using plasma proteomic profiles from 53,014 participants in the UK Biobank Pharma Proteomics Project, we pretrained the encoder by masked-protein reconstruction and evaluated whether disease models developed from comprehensive 2,920-protein profiles could be reused with a predefined subset of 1,460 proteins. Across 144 diseases, median AUC was 0.679 with comprehensive coverage and 0.637 when the same encoder and disease models were applied to partial-coverage representations without refitting; retraining only the disease-specific models increased median AUC to 0.673. Under partial coverage, protein-token proteomic risk scores (ProRS) exceeded coefficient-truncated LASSO ProRS by a median paired AUC difference of 0.027 and were comparable to LASSO ProRS refitted using outcome labels, with a median difference of 0.003. Performance was relatively stable across cardiovascular-kidney-metabolic diseases but more heterogeneous across autoimmune diseases, while protein-token representations improved discrimination beyond clinical covariates for 10 of 12 focused diseases under partial coverage without refitting. These results support self-supervised protein-token representations as a strategy for building proteomic prediction models that remain usable across heterogeneous measurement settings.
Related
- Large-scale plasma proteomic profiling enhances prediction of heart failure among hypertensive individuals: prospective insights from the UK Biobank.
- Plasma proteomic profiles associated with low-dose computed tomography findings and incident lung cancer: a prospective pilot study
- Large-scale proteomic characterization reveals clinically relevant subtypes of early-stage papillary thyroid cancer for prognostic prediction
- Differentiation of Human Glioblastoma Stem Cells as a Prospective Tool to Find New Biomarkers and Molecular Targets: A Proteomic and Lipidomic Approach
- Battery performance prediction using supervised learning: a comparative study
BibTeX
@misc{nam2026selfsupervised,
title = {{Self-supervised plasma proteomic representations for prospective disease prediction across varying protein availability}},
author = {Y. Nam and T. M. Westbrook and J. Woerner and J. Joo and M. E. Lee and M. McKeague and A. E. Baxter and Shwetank and M. Ionita and S. A. Apostolidis and A. R. Greenplate and E. J. Wherry and D. Kim},
howpublished = {medRxiv},
year = {2026},
month = sep,
doi = {10.64898/2026.09.10.26362789},
url = {https://www.medrxiv.org/content/10.64898/2026.09.10.26362789v1},
note = {CC BY-NC-ND},
}
RIS
TY - UNPB
TI - Self-supervised plasma proteomic representations for prospective disease prediction across varying protein availability
AU - Y. Nam
AU - T. M. Westbrook
AU - J. Woerner
AU - J. Joo
AU - M. E. Lee
AU - M. McKeague
AU - A. E. Baxter
AU - Shwetank
AU - M. Ionita
AU - S. A. Apostolidis
AU - A. R. Greenplate
AU - E. J. Wherry
AU - D. Kim
PY - 2026
DA - 2026/09/10
PB - medRxiv
DO - 10.64898/2026.09.10.26362789
UR - https://www.medrxiv.org/content/10.64898/2026.09.10.26362789v1
AB - Large-scale plasma proteomics offers opportunities to characterize disease susceptibility and improve prospective risk prediction, but transferring proteomic predictors across datasets remains challenging because measured protein sets differ across cohorts, study phases and assay configurations. Here we developed a self-supervised protein-token Transformer that maps the proteins observed in each sample to a fixed-dimensional participant representation, allowing unavailable proteins to be omitted rather than imputed. Using plasma proteomic profiles from 53,014 participants in the UK Biobank Pharma Proteomics Project, we pretrained the encoder by masked-protein reconstruction and evaluated whether disease models developed from comprehensive 2,920-protein profiles could be reused with a predefined subset of 1,460 proteins. Across 144 diseases, median AUC was 0.679 with comprehensive coverage and 0.637 when the same encoder and disease models were applied to partial-coverage representations without refitting; retraining only the disease-specific models increased median AUC to 0.673. Under partial coverage, protein-token proteomic risk scores (ProRS) exceeded coefficient-truncated LASSO ProRS by a median paired AUC difference of 0.027 and were comparable to LASSO ProRS refitted using outcome labels, with a median difference of 0.003. Performance was relatively stable across cardiovascular-kidney-metabolic diseases but more heterogeneous across autoimmune diseases, while protein-token representations improved discrimination beyond clinical covariates for 10 of 12 focused diseases under partial coverage without refitting. These results support self-supervised protein-token representations as a strategy for building proteomic prediction models that remain usable across heterogeneous measurement settings.
ER -