The C-score: a Bayesian framework to sharply improve proteoform scoring in high-throughput top down proteomics

Area of Science:

• Proteomics
• Biochemistry
• Computational Biology

Background:

• Automated analysis of top-down proteomics data requires enhanced scoring for accurate protein identification.
• Distinguishing highly related protein forms (proteoforms) remains a challenge in current proteomic analyses.

Purpose of the Study:

• To introduce the C-score (Characterization Score), a novel Bayesian approach for improved proteoform identification and characterization.
• To integrate expert knowledge into generative models for top-down proteomics data analysis.

Main Methods:

• Developed a Bayesian framework (C-score) incorporating protein properties and analytical system characteristics.
• Implemented intelligent weighting for site-specific modifications and accounted for fragmentation propensities and errors.
• Compared C-score performance against existing probability-based scoring systems (ProSightPC, ProSightPTM) using a curated dataset.

Main Results:

• The C-score framework demonstrated a marked improvement in scoring performance.
• Achieved a significantly higher area under the curve (AUC) of 0.99 compared to 0.78 for existing methods.
• Validated performance on a manually curated set of 295 human proteoforms.

Conclusions:

• The C-score offers a substantial advancement for proteoform identification and characterization in top-down proteomics.
• The Bayesian approach effectively leverages expert knowledge for more accurate data interpretation.
• This scoring system has the potential to improve the automated processing of complex proteomic datasets.