Quantifying and correcting slide-to-slide variation in multiplexed immunofluorescence images

Area of Science:

• Single-cell biology
• Computational pathology
• Bioinformatics

Background:

• Multiplexed imaging is a powerful single-cell assay for studying cell-cell interactions.
• The complex data structure of multiplexed imaging is prone to technical variability, hindering reliable inference.
• Standardized processing and normalization techniques for multiplexed imaging data are currently limited.

Purpose of the Study:

• To implement and compare various data transformation and normalization algorithms for multiplexed imaging data.
• To adapt existing methods like ComBat and functional data registration for slide effect removal in this domain.
• To establish an evaluation framework for comparing normalization approaches in multiplexed imaging.

Main Methods:

• Adaptation of ComBat and functional data registration methods for slide effect normalization.
• Development of an evaluation framework to assess normalization performance.
• Comparison of various normalization techniques on multiplexed imaging datasets.

Main Results:

• Demonstration of significant slide-to-slide variation in raw multiplexed imaging data.
• Validation that several normalization methods reduce technical variation while preserving/enhancing biological signals.
• Identification of slide mean division and functional data registration as top-performing methods under the evaluation framework.

Conclusions:

• The proposed normalization approaches improve data quality for multiplexed imaging.
• The developed evaluation framework provides robust criteria for assessing normalization methods.
• This work lays the groundwork for enhanced data processing and analysis in multiplexed imaging studies.