Workflow for Evaluating Normalization Tools for Omics Data Using Supervised and Unsupervised Machine Learning.

Area of Science:

• Analytical Chemistry
• Biotechnology
• Data Science

Background:

• High-quality omics results necessitate addressing systematic variability in mass spectrometry (MS) data.
• Effective data normalization is crucial for minimizing this variability.
• Existing open-source software for normalization strategy selection has limitations, particularly for users exploring novel approaches.

Purpose of the Study:

• To present a straightforward workflow for identifying optimal normalization strategies in MS data.
• To enable the evaluation of any normalization strategy for any MS data type.
• To incorporate both supervised and unsupervised machine learning for performance assessment.

Main Methods:

• Development of a "DIY" workflow for evaluating MS data normalization strategies.
• Utilization of straightforward evaluation metrics.
• Application of supervised and unsupervised machine learning techniques.

Main Results:

• Demonstrated the workflow's utility on two distinct datasets: ESI-MS of lipids from latent fingerprints and MALDI-MSI of metabolites from cancer spheroids.
• Successfully identified the best-performing normalization strategies for each dataset.
• Validated the flexibility and effectiveness of the proposed workflow.

Conclusions:

• The developed workflow provides a simple yet powerful tool for optimizing MS data normalization.
• It empowers researchers to test and validate novel normalization strategies beyond existing software limitations.
• Facilitates the achievement of high-quality omics data through tailored normalization approaches.