Joint-RPCA for Multi-omics Integration
Cross-language validation of R/Bioconductor and Python workflows for microbiome data.
Overview
This active research and software-development project investigates Joint Robust Principal Component Analysis (Joint-RPCA) for integrating microbiome multi-omics data. The practical goal is to make results from R/Bioconductor and Python/Gemelli workflows comparable by controlling preprocessing, configuration, randomness, and output alignment.
Methodological focus
- Apply equivalent abundance, prevalence, and frequency filtering before ordination.
- Preserve a shared sample set across multiple assays.
- Run both implementations with fixed configuration and random seeds.
- Compare sample scores, feature loadings, explained variation, and cross-validation outputs.
- Account for sign and rotational indeterminacy when comparing latent representations.
Validation evidence
The working IBD multi-omics dataset contains 60 shared samples. Before frequency filtering, the current workflow retains 459 metagenomic features and 800 metatranscriptomic features. A 10% frequency threshold retains 334 and 785 features, respectively. These counts are treated as reproducibility checks rather than biological conclusions.
Engineering contribution
The work includes test coverage for input filtering, reproducible Quarto documentation, structured configuration files, and interoperable R/Python outputs. Negative or missing values are rejected before transformation, and threshold semantics are tested explicitly.
Limitations
- Equivalent software settings do not guarantee identical numerical optimization paths.
- Latent components may require sign or orthogonal alignment before comparison.
- Biological interpretation requires domain validation beyond computational agreement.
Repositories
Tools: R, Python, Bioconductor, mia, MultiAssayExperiment, Gemelli, Quarto, GitHub Actions