Alpha-XIC

Alpha-XIC improves peptide identification in data-independent acquisition (DIA) mass spectrometry by using a deep neural network to compute robust coelution scores for peak groups.

Key Features:

• Neural Network-Based Coelution Scoring: Alpha-XIC uses a deep neural network to learn implicit coelution features from identified peptide data, capturing shape similarity and retention time relationships.
• Robustness in Interference Conditions: The model minimizes the impact of noise and interference to provide reliable coelution scores for peak groups affected by interference.
• Integration with Existing Identification Engines: Alpha-XIC appends its coelution scores to outputs from existing identification engines to support statistical validation and correction of misidentified peptides.

Scientific Applications:

• HeLa Dataset Evaluation: On HeLa datasets with gradient lengths from 0.5 h to 2 h, Alpha-XIC increased the number of identified precursors by 16.7% to 49.1% at 1% FDR.
• LFQbench Dataset Testing: On mixed-species LFQbench samples with known ratios, Alpha-XIC increased the number of peptides and proteins within valid ratios by up to 16.6% and 13.8%, respectively.

Methodology:

Alpha-XIC trains a deep neural network on identified peptide data to derive coelution characteristics and generates coelution scores that are appended to identification engine outputs for downstream statistical validation.