LABEL-FREE TARGETED PROTEOMICS DATA ANALYSIS WORKFLOW SELECTION - BENCHMARKING AI-BASED AND DATA-DRIVEN APPROACHES

Area of Science:

• Proteomics
• Quantitative Biology
• Biomarker Discovery

Background:

• Global proteomics identifies more proteins, but quantitative performance needs optimization.
• Label-free targeted proteomics offers a cost-efficient validation method.
• Limited systematic evaluations exist for label-free targeted proteomics workflows, especially with AI tools.

Purpose of the Study:

• Benchmark various data analysis workflows for label-free targeted proteomics.
• Evaluate missing-data imputation (MDI) and data consolidation strategies.
• Establish a validation framework for global proteomics results.

Main Methods:

• Compared MDI strategies: no MDI, k-nearest neighbors, data-driven, MSstats, AI-based.
• Tested consolidation/testing frameworks: summation, best-peak, AI scaling, p-value integration, MSstats TMP, multivariate testing.
• Utilized a controlled dataset with spiked yeast proteins in a human background.

Main Results:

• Data-driven MDI combined with p-value integration demonstrated superior performance.
• This approach outperformed others in accuracy, precision, specificity, and false-discovery rate.
• Careful workflow selection significantly improves quantitative outcomes over simple summation.

Conclusions:

• Data-driven MDI and p-value integration provide a robust workflow for label-free targeted proteomics.
• This optimized approach enhances quantitative results for biomarker validation.
• The findings offer a generally applicable default workflow for label-free biomarker validation.