CoDA-Aware Feature Selection for Biomarker Panels

Feature-selection methods tailored to compositional omics data. The exported functions return per-feature importance scores and selected features, while the accompanying `mlr3filters::Filter` classes expose the methods inside the existing `OmicPipeline` feature-selection path.

The implemented selectors cover five complementary CoDA-aware strategies: - `codaFS_plr_variance()`: pairwise log-ratio screening with feature aggregation. - `codaFS_selbal_wrapper()`: sparse forward-balance selection using the official `selbal` package when available, or a native fallback. - `codaFS_codacore_wrapper()`: sparse balance selection via CoDaCoRe using the existing OmicSelector interface. - `codaFS_logcontrast_lasso()`: CLR-space lasso with post-fit zero-sum projection. - `codaFS_stability_logratio()`: stability selection on pairwise log-ratio features with elastic-net aggregation back to individual miRNAs.

References

Aitchison J. (1986). The Statistical Analysis of Compositional Data. Chapman and Hall.

Rivera-Pinto J, Egozcue JJ, Pawlowsky-Glahn V, et al. (2018). Balances: a new perspective for microbiome analysis. mSystems, 3(4), e00053-18.

Gordon-Rodriguez E, Susin A, McEwen JD, et al. (2022). CoDaCoRe: learning sparse log-ratios for high-throughput sequencing data. Bioinformatics, 38(12), 3179-3186.

Lin W, Shi P, Feng R, Li H. (2014). Variable selection in regression with compositional covariates. Biometrika, 101(4), 785-797.

Meinshausen N, Buhlmann P. (2010). Stability selection. Journal of the Royal Statistical Society Series B, 72(4), 417-473.