BatchI: Batch effect Identification in high-throughput screening data using a dynamic programming algorithm.

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Batch effects are a significant source of bias in high-throughput biological experiments.
• Existing methods for batch effect removal often require prior knowledge of batch assignments.
• Few methods can identify batch effects of unknown structure.

Purpose of the Study:

• To introduce an original algorithm, BatchI, for identifying batch effects in omics data using dynamic programming.
• To address the challenge of batch effect identification when batch structure is unknown.
• To provide a robust method for handling bias in biological measurements.

Main Methods:

• BatchI algorithm partitions high-throughput experiment samples into sub-series representing estimated batches.
• Dynamic programming is employed to maximize dispersion between batches while minimizing dispersion within batches.
• The algorithm was tested on datasets with and without prior batch information.

Main Results:

• BatchI accurately identified known batch divisions in datasets with prior information.
• The algorithm improved data quality and biological information content in datasets with unknown batch structures.
• Batch effect correction, guided by BatchI, resulted in higher intra-group correlations.

Conclusions:

• BatchI is an effective tool for identifying unknown batch effects in omics data.
• Accurate batch effect control is crucial for maximizing discoveries in biological experiments.
• The algorithm serves as a valuable starting point for batch effect correction methods.