Protein Abundance Inference via Expectation Maximization in Fluorosequencing.

Area of Science:

• Proteomics
• Computational Biology
• Biotechnology

Background:

• Fluorosequencing generates vast single-peptide data, but lacks a robust method for quantitative protein abundance estimation.
• Existing peptide classification tools provide valuable input but require integration into a protein-level quantification strategy.

Purpose of the Study:

• To introduce a probabilistic framework for accurate protein abundance estimation from fluorosequencing data.
• To develop a scalable computational method bridging peptide classification and protein quantification.

Main Methods:

• Implemented an expectation-maximization (EM) based probabilistic framework to estimate relative protein abundances.
• Integrated results from existing peptide-classification tools into the EM algorithm.
• Assessed performance using simulated five-protein mixtures and large-scale human proteome simulations.

Main Results:

• The EM-based framework significantly reduced mean absolute error in relative abundance compared to a uniform-abundance guess.
• The method demonstrated scalability, processing millions of reads efficiently on standard and high-performance computing systems.
• Accuracy improvements were more pronounced with lower fluorosequencing error rates, suggesting potential for future advancements.

Conclusions:

• EM-based inference provides a scalable, model-driven solution for quantitative proteomics using fluorosequencing data.
• The framework enhances protein abundance estimates and can serve as a refinement step for other inference methods.
• Improvements in fluorosequencing technology combined with this computational approach promise substantially more accurate high-throughput single-molecule proteomics.