Trend-Aligner: A Retention Time Modeling-Based Feature Alignment Method for Untargeted LC-MS Data Analysis.

Area of Science:

• Analytical Chemistry
• Biochemistry
• Computational Biology

Background:

• Accurate cross-sample quantification in LC-MS multiomics relies on matching analytes across runs.
• Retention time shifts between runs necessitate dedicated alignment algorithms.
• Existing algorithms often overlook chromatographic principles governing retention time shifts.

Purpose of the Study:

• Introduce Trend-Aligner, a novel feature alignment algorithm for untargeted MS data.
• Explicitly model retention time shifts using chromatographic principles for analyte-wise correction.
• Enhance the accuracy and reliability of multiomics data analysis.

Main Methods:

• Decompose retention time shifts into global (nonlinear warping) and local (latent factor model) components.
• Model global shifts capturing systematic variations in chromatographic conditions.
• Model local shifts accounting for analyte-specific responses to chromatographic changes.

Main Results:

• Trend-Aligner demonstrated superior accuracy compared to 11 existing alignment algorithms across diverse metabolomic and proteomic datasets.
• Achieved high sensitivity and specificity in feature alignment.
• Showcased an 82.5% increase in identified peptides post-MBR integration compared to MaxQuant.

Conclusions:

• Trend-Aligner provides a principled approach to retention time alignment in LC-MS multiomics.
• The algorithm enhances feature identification and quantification accuracy.
• Trend-Aligner offers a significant improvement for real-world multiomics applications.