Removing batch effects from purified plasma cell gene expression microarrays with modified ComBat.

Area of Science:

• Genomics and Bioinformatics
• Computational Biology
• Biostatistics

Background:

• Gene expression profiling (GEP) using microarray analysis is crucial for clinical risk assessment and diagnostics.
• Batch effects from sample preparation and scanners introduce systematic variability, complicating long-term studies and meta-analyses.
• ComBat is an existing method to mitigate batch effects in microarray data.

Purpose of the Study:

• To introduce M-ComBat, a modified ComBat algorithm for gene expression data.
• M-ComBat centers data to the location and scale of a pre-determined 'gold-standard' batch.
• To enhance the performance of predictive models in meta-analysis by adjusting for batch effects.

Main Methods:

• Applied ComBat and M-ComBat to combined microarray datasets with known batch variations.
• Utilized external datasets from HOVON-65/GMMG-HD4 and MRC-IX trials.
• Compared fixed and validated gene risk signatures developed on a 'gold-standard' batch.

Main Results:

• Both ComBat and M-ComBat successfully eliminated systematic batch effects across all probes.
• M-ComBat demonstrated superior agreement with the 'gold-standard' batch in terms of risk score distribution and high-risk subject identification.
• M-ComBat improved the performance of risk scores, enabling more precise identification of high-risk cohorts.

Conclusions:

• M-ComBat is a practical enhancement to ComBat, offering improved control over data location and scale after batch effect adjustment.
• This method ensures historical predictive models remain effective on future gene expression data despite systematic changes.
• M-ComBat facilitates more reliable and powerful batch-effect adjusted data analysis for meta-analysis and clinical applications.