DeepRTAlign: toward accurate retention time alignment for large cohort mass spectrometry data analysis

Area of Science:

• Biochemistry
• Computational Biology
• Analytical Chemistry

Background:

• Retention time (RT) alignment is critical for large-scale liquid chromatography-mass spectrometry (LC-MS) proteomic and metabolomic studies.
• Current alignment methods struggle with simultaneous monotonic and non-monotonic RT shifts.
• RT alignment is a significant bottleneck in LC-MS data analysis.

Purpose of the Study:

• To develop a deep learning-based tool, DeepRTAlign, for robust RT alignment in large cohort LC-MS data.
• To address the limitations of existing methods in handling complex RT shifts.
• To improve the accuracy and sensitivity of proteomic and metabolomic data analysis.

Main Methods:

• Development of DeepRTAlign, a novel deep learning algorithm for RT alignment.
• Benchmarking DeepRTAlign against state-of-the-art alignment tools.
• Validation on diverse real-world and simulated proteomic and metabolomic datasets.
• Application of aligned MS features for biomarker discovery.

Main Results:

• DeepRTAlign demonstrated superior performance compared to existing methods.
• The tool effectively handles both monotonic and non-monotonic RT shifts.
• Improved identification sensitivity without compromising quantitative accuracy was observed.
• A robust classifier for early hepatocellular carcinoma recurrence prediction was developed using DeepRTAlign-aligned data.

Conclusions:

• DeepRTAlign offers an advanced solution for RT alignment in large cohort LC-MS studies.
• The tool enhances data analysis and facilitates biomarker discovery in proteomics and metabolomics.
• DeepRTAlign addresses a major challenge in the field, paving the way for more reliable LC-MS data interpretation.