CCSbase

CCSbase predicts collision cross section (CCS) values and supports identification of unknown compounds in ion mobility–mass spectrometry (IM-MS) and large-scale untargeted analyses such as metabolomics and drug metabolite identification.

Key Features:

• IM-MS and CCS utilization: Leverages ion mobility–mass spectrometry (IM-MS) and collision cross section (CCS) as a two-dimensional separation and structural descriptor for compound identification.
• Comprehensive CCS database: Aggregates high-quality reference CCS values spanning a wide chemical space.
• Machine learning (ML) prediction: Employs machine learning models trained on the reference CCS collection to predict CCS values.
• MQN-based descriptors: Identifies molecular quantum numbers (MQNs) as structural characteristics that contribute to CCS variance.
• Unsupervised MQN clustering: Applies unsupervised clustering on MQNs to define chemically coherent clusters for modeling.
• Cluster-specific models: Trains separate ML prediction models for each MQN-derived cluster, improving accuracy relative to a single global model.
• Model characterization: Models are robustly trained and characterized to achieve high accuracy across diverse chemical structures.

Scientific Applications:

• Untargeted metabolomics: Supports identification of unknown metabolites in large-scale untargeted metabolomics using IM-MS CCS data.
• Drug metabolite identification: Aids identification of drug metabolites by providing reference and predicted CCS values for structural discrimination.
• IM-MS compound identification: Enhances structural elucidation and candidate filtering in IM-MS-based compound identification workflows.

Methodology:

Collects a comprehensive set of reference CCS values, computes molecular quantum numbers (MQNs), performs unsupervised clustering on MQNs, trains machine learning (ML) prediction models for each MQN-derived cluster, and evaluates model performance against a single global model.