Improving cross-study prediction through addon batch effect adjustment or addon normalization.

Area of Science:

• Bioinformatics
• Computational Biology
• Biostatistics

Background:

• Medical tests using high-dimensional molecular data often exhibit poor performance on external datasets.
• This prediction error is higher than internal validation suggests, limiting clinical adoption.
• Systematic differences between datasets contribute to this performance gap.

Purpose of the Study:

• To investigate addon normalization and batch effect removal techniques for improving cross-study prediction performance.
• To reduce systematic differences between external and original datasets for molecular data classification.
• To enhance the clinical utility of molecular data-derived medical tests.

Main Methods:

• Evaluation of addon normalization and seven batch effect removal methods.
• Application to a large collection of microarray gene expression datasets.
• Assessment of impact on cross-study prediction performance for common classifiers.

Main Results:

• Some addon normalization and batch effect removal techniques significantly reduce prediction error on external data.
• The effectiveness of these techniques varies across different methods and classifiers.
• Demonstrated improvement in cross-study prediction performance.

Conclusions:

• Addon normalization and batch effect removal are effective strategies to improve the generalizability of molecular data-based medical tests.
• These methods can help bridge the gap between internal validation and real-world clinical performance.
• The R package 'bapred' provides implementation of the investigated addon methods.