omicverse-bulk-metabol-preprocessing

You can preprocess metabolomics peak tables into analysis-ready AnnData objects by applying deterministic loading, imputation, normalization, transformation, and batch correction. This pipeline supports MetaboAnalyst CSV, generic wide CSV, and LC-MS workflows.

Handling missing values in metabolomics data requires imputation methods like qrilc, knn, half_min, or zero imputation. These options address different missingness patterns before applying normalization and transformation to the peak table.

Batch correction and drift correction for LC-MS metabolomics datasets work by applying ComBat correction, SERRF, and sample-level QC. These methods handle injection-order drift and batch variation when QC pools are present in the run.

You can normalize and transform metabolomics features in one pipeline using pqn, tic, median, or mstus normalization, followed by log, glog, pareto, or autoscale transformation. The chainable execution applies these steps sequentially to the data.

This metabolomics preprocessing workflow supports MetaboAnalyst CSV format, generic wide CSV, and LC-MS peak tables. It reads these formats into a consistent AnnData structure for downstream differential or multivariate analysis.

You need sample QC and batch correction before downstream metabolomics analysis to remove injection-order drift and batch variation from LC-MS datasets. Correcting these artifacts ensures the resulting matrix is clean for accurate differential or multivariate modeling.