CPred: Charge State Prediction for Modified and Unmodified Peptides in Electrospray Ionization

Area of Science:

• Proteomics
• Analytical Chemistry
• Computational Biology

Background:

• Mass spectrometry is vital for peptide identification, relying on the mass-to-charge ratio.
• Peptide charge state determination is complex, especially with multiply charged ions from electrospray ionization.

Purpose of the Study:

• To develop a neural network (CPred) for accurate prediction of peptide charge state distributions.
• To enhance confidence in peptide identification through improved charge state prediction.

Main Methods:

• Developed CPred, a neural network model.
• Trained CPred on large-scale synthetic data including tryptic/non-tryptic peptides and various fragmentation methods.
• Evaluated CPred on independent test datasets using Pearson correlation coefficient.

Main Results:

• CPred accurately predicts charge state distributions (+1 to +7) for modified and unmodified peptides.
• Achieved high correlations (up to 0.9997117) between predicted and acquired charge state distributions.
• Identified the importance of modifications and peptide properties in proton affinity.

Conclusions:

• CPred offers accurate charge state distribution predictions, significantly aiding peptide identification.
• The model's ability to incorporate modifications improves its predictive power.
• CPred serves as a valuable new feature for rescoring peptide identifications in mass spectrometry.