AART enables fast and accurate cross-platform proteomic translation

Area of Science:

• Proteomics
• Biomarker Discovery
• Data Science

Background:

• Plasma proteomic profiling is crucial for biomarker discovery, disease prediction, and patient stratification.
• Technical variability across assay platforms (e.g., Olink, SomaScan, mass spectrometry) hinders reproducibility, data integration, and model transferability.

Purpose of the Study:

• To introduce AART, a novel cross-platform proteomic translation framework.
• To enhance the accuracy, reproducibility, and scalability of proteomic data analysis across diverse platforms.

Main Methods:

• AART integrates matched-protein ridge regression with proteome-wide residual learning.
• Benchmarked across three independent cohorts and three major platforms (Olink, SomaScan, mass spectrometry).

Main Results:

• AART significantly outperformed baseline methods in cross-platform proteomic translation, showing an average improvement of 92.0% over direct mapping.
• AART improved the reproducibility of association analyses for type 2 diabetes and Alzheimer's disease by 75.5% and 370.6%, respectively.
• AART enhanced diagnostic accuracy for amyotrophic lateral sclerosis by 92.6% and was orders of magnitude faster than existing methods.

Conclusions:

• AART provides a fast, accurate, and scalable solution for cross-platform proteomic translation.
• The framework facilitates more reproducible, transferable, and integrated proteomic research, enabling large-scale biobank applications.