Collisional Cross-Section Prediction for Multiconformational Peptide Ions with IM2Deep

Area of Science:

• Analytical Chemistry
• Computational Chemistry
• Biochemistry

Background:

• Peptide collisional cross-section (CCS) prediction is complex due to multiple gas-phase conformations.
• This complexity hinders downstream proteomics data analysis, including identification and quantification.
• Accurate CCS values are crucial for interpreting mass spectrometry data.

Purpose of the Study:

• To develop an improved IM2Deep model for accurate prediction of peptide collisional cross-section (CCS) values.
• To address the challenge of multiconformational peptide ions in gas-phase experiments.
• To enhance the utility of CCS prediction in proteomics.

Main Methods:

• Training an enhanced IM2Deep model on a curated dataset of publicly available data.
• Utilizing a large and comprehensive dataset for model training.
• Employing a tailored neural network architecture for CCS prediction.

Main Results:

• The enhanced IM2Deep model accurately predicts CCS values for multiconformational peptides.
• The model demonstrates high precision even for peptides with a single observed conformation.
• The improved prediction capabilities facilitate more reliable proteomics data analysis.

Conclusions:

• The enhanced IM2Deep model offers a significant advancement in predicting peptide CCS values.
• Accurate CCS prediction of multiple peptide conformational states is now achievable.
• The open-source availability of IM2Deep promotes wider adoption and research in proteomics.