# Fast feature (m/z) selection using multiple hypothesis testing (LIMMA/ANOVA/Kruskal) with optional class balancing (no/up/down/SMOTE) Runs univariate statistical tests across all features (columns) to identify discriminant m/z values. Supports three tests: * "Limma": moderated F-test via the limma R package on k-1 contrasts between groups * "anova": classical one-way ANOVA (implemented in C++ for speed) * "kruskal": Kruskal–Wallis rank-sum test (implemented in C++ with tie correction) ```r SelectionVarStat( X, Y, stat.test = c("Limma", "anova", "kruskal"), pi0.method = "abh", fdr = 0.05, Sampling = c("no", "up", "down", "smote"), seed = NULL ) ``` ## Arguments - `X`: Numeric matrix with samples in rows and features (peaks) in columns. If a sparse Matrix is provided, it is coerced to a base R dense matrix. Infinities are set to NA; NAs are allowed. - `Y`: Class labels. A factor (or coercible to factor) of length `nrow(X)`. Must contain at least two levels. - `stat.test`: Character string; which univariate test to use. One of `"Limma"`, `"anova"`, or `"kruskal"`. Default is `"Limma"`. * Limma uses `limma::lmFit` on `~ 0 + Y`, applies k-1 contrasts versus a reference group with `limma::makeContrasts` and then `limma::eBayes` to compute a moderated F p-value testing overall between-group differences. * anova uses a fast C++ one-way ANOVA (upper-tail F p-value). * kruskal uses a fast C++ Kruskal–Wallis test with tie correction (upper-tail chi-square p-value). - `pi0.method`: Character; method for `cp4p::estim.pi0` and `cp4p::adjust.p`. Default `"abh"`. See cp4p documentation for options. - `fdr`: Numeric in (0, 1]; FDR threshold used to select features after p-value adjustment. Default `0.05`. - `Sampling`: Character string; optional class balancing applied before testing. One of `"no"`, `"up"`, `"down"`, `"smote"`. Default `"no"`. * "up": up-samples minority classes to the majority size (base R implementation). * "down": down-samples majority classes to the minority size (base R). * "smote": uses the package’s internal `smote_classif()` on `data.frame(Y, X)`. Column names of `X` are preserved; if SMOTE fails or yields \<2 classes, the function falls back to up-sampling. - `seed`: Optional integer. If provided, sets the random seed for up/down sampling and SMOTE to ensure reproducibility. ## Returns A list with: * `nb_to_sel`: integer, `floor(ncol(X) * (1 - pi0))` using `cp4p::estim.pi0`. * `n_selected_fdr`: integer, number of features with adjusted p-value \< `fdr`. * `sel_moz`: character vector of selected feature names (columns of `X`) with adjusted p-value \< `fdr`. * `ap`: the full object returned by `cp4p::adjust.p` (contains adjusted p-values). ## Details Optional class balancing can be applied before testing: no sampling, up-sampling, down-sampling, or SMOTE using the internal `smote_classif()` function. P-values are adjusted with the cp4p R package, and features below the FDR threshold are returned, together with an estimated proportion of nulls, `pi0`. Missing values are preserved as `NA`: for ANOVA/Kruskal, the C++ backends skip NAs per feature; for Limma, missing values are handled by limma internally. Non-finite values (Inf/-Inf) are coerced to `NA`. * Limma design: `model.matrix(~ 0 + Y)` is used; a full-rank set of k-1 contrasts vs the first level is constructed with `limma::makeContrasts`. The returned p-values are moderated F-test p-values for overall group differences. * ANOVA/Kruskal implementations are in C++ (see `anova_cols_cpp` and `kruskal_cols_cpp`); they process all features in one pass and skip NAs per feature. * Non-finite p-values (if any) are set to 1 before `pi0`/adjustment. * SMOTE requires at least two observations per class; otherwise the function automatically falls back to up-sampling. * For limma, if residual degrees of freedom are not positive (`nrow(X) - nlevels(Y) \<= 0`), p-values are set to 1 with a warning. ## Examples ```r ############################################################################### ## 1. Pre-processing of mass spectra # load mass spectra and their metadata data("CitrobacterRKIspectra","CitrobacterRKImetadata", package = "MSclassifR") # standard pre-processing of mass spectra spectra <- MSclassifR::SignalProcessing(CitrobacterRKIspectra) # detection of peaks in pre-processed mass spectra peaks <- MSclassifR::PeakDetection(x = spectra, labels = CitrobacterRKImetadata$Strain_name_spot) # build matrix with intensities of peaks (rows = samples, columns = m/z) Y <- factor(CitrobacterRKImetadata$Species) xy <- build_XY_from_peaks(peaks, labels = Y, normalize = "max", sparse = FALSE) X <- xy$X Y <- xy$Y ############################################################################### ## 2. Estimate the optimal number of peaks to discriminate the different species OptiPeaks <- MSclassifR::SelectionVarStat(X, Y, stat.test = "Limma", pi0.method = "abh", fdr = 0.05, Sampling = "smote", seed = 1) ## Estimation of the optimal number of peaks to discriminate species (from the pi0 parameter) OptiPeaks$nb_to_sel ## discriminant mass-to-charge values estimated using a 5 per cent false discovery rate OptiPeaks$sel_moz ## p-values and adjusted p-values estimated for all the tested mass-to-charge values OptiPeaks$ap$adjp ``` ## See Also limma::lmFit, limma::contrasts.fit, limma::eBayes, limma::makeContrasts; cp4p::estim.pi0, cp4p::adjust.p

SelectionVarStat function