Assessing the impact of batch effect associated missing values on downstream analysis in high-throughput biomedical...

Area of Science:

• Biostatistics
• Bioinformatics
• Computational Biology

Background:

• Batch effect associated missing values (BEAMs) arise from integrating multi-batch omics data with varying feature coverage.
• BEAMs pose significant challenges for downstream data analysis, including missing value imputation (MVI) and batch effect correction (BE).

Purpose of the Study:

• To investigate the impact of BEAMs on the performance of various MVI and BE correction algorithms (BECAs).
• To evaluate the robustness of commonly used MVI methods when confronted with BEAMs in real-world and simulated datasets.

Main Methods:

• Simulations and analysis of real-world datasets (e.g., CPTAC).
• Evaluation of six MVI methods (KNN, Mean, MinProb, SVD, MICE, RF) alongside two BECAs (ComBat, limma).

Main Results:

• BEAMs significantly degrade MVI performance, leading to inaccurate imputations and inflated P-values.
• KNN, SVD, and RF methods propagated random signals, causing false statistical confidence.
• Mean and MinProb imputations introduced artifacts, with detrimental effects increasing with BEAM severity.

Conclusions:

• BEAMs necessitate comprehensive assessments and tailored strategies for reliable analysis of multi-batch omics data.
• Current MVI and BE correction methods may not adequately handle BEAMs, compromising data integrity.
• Future research should focus on advanced simulations and dedicated MVI methods to address BEAMs effectively.