gly_wgcna(), gly_wgcna_(), gly_consensus_clustering(), and gly_consensus_clustering_() are removed. These functions were deprecated in 0.7.0. Use the WGCNA and ConsensusClusterPlus packages directly.gly_enrich_go(), gly_enrich_kegg(), gly_enrich_reactome(), gly_enrich_wikipathways(), gly_enrich_do(), gly_enrich_ncg(), and gly_enrich_dgn(). Use the corresponding functions in the glyfun package instead.get_tidy_result() is now type-stable: it always returns a tibble (or errors), instead of sometimes returning a list when which is omitted. When tidy_result is a list, the which argument is now required (#9).gly_ttest(), gly_wilcox(), gly_anova(), and gly_kruskal(). Effect sizes (Cohen's d, rank-biserial correlation, eta-squared, and epsilon-squared) are now included in the tidy results (#8).gly_ttest() and gly_wilcox().estimate sign to be inconsistent in gly_kruskal() results.meta_data to all glystats results. This allows storing additional information with the result object that can be accessed later (#7).gly_ttest() output.gly_enrich_go(), gly_enrich_kegg(), gly_enrich_reactome()). These functions now use explicit parameters instead of .... Users must update their code to use new parameters (e.g., p_cutoff instead of pvalueCutoff).gly_wgcna(), gly_wgcna_(), gly_consensus_clustering(), and gly_consensus_clustering_() are deprecated and will be removed in a future version. These functions are too interactive and complex for a pipeline-friendly package.gly_enrich_wikipathways() and gly_enrich_wikipathways_() for WikiPathways enrichment analysis.gly_enrich_do() and gly_enrich_do_() for Disease Ontology enrichment analysis.gly_enrich_ncg() and gly_enrich_ncg_() for Network of Cancer Genes enrichment analysis.gly_enrich_dgn() and gly_enrich_dgn_() for DisGeNET enrichment analysis.gly_hclust() leaves a plotting device unclosed.gly_limma() fails for site-specific motif experiments.subject_cols argument in gly_limma() has been renamed into subject_col.gly_limma() fails to work with experiments with all-NA variables.gly_ancova() for Analysis of Covariance (ANCOVA).gly_limma() and gly_limma_() through the covariate_cols and the covariates arguments.gly_limma() by default. This is more appropriate for glycomics and glycoproteomics data.gly_roc().filter_sig_vars() now has a default fc_cutoff of 2 for glycoproteomics data, and NULL for other types of data.gly_pca() fails to work with experiments with constant variables. Now the constant variables are removed before PCA, with a warning issued.gly_limma() is not correct for glycomics data.universe argument of gly_enrich_go(), gly_enrich_kegg(), and gly_enrich_reactome() accepts a glyexp::experiment() object now.gly_limma().sample or variable column in the results now, not appending to the end as previously.filter_sig_vars() to inform about using the same experiment used for DEA.filter_sig_vars() now doesn't require comparison to be provided for gly_limma() results. If comparison is not provided, a variable will be kept if it is significant in any comparison.... to the underlying statistical functions.glyexp 0.10.0.This is a big update! We make some breaking changes to the API, mainly DEA functions. Now you can ensure these properties for all DEA results:
post_hoc column in tidy_result$main_test of results from gly_anova() and gly_kruskal() is now in the format of "ref_vs_test" instead of "ref-test".group1 and group2 columns in tidy_result$post_hoc_test of results from gly_anova() and gly_kruskal() are renamed into ref_group and test_group for more clarity.tidy_result of gly_limma(), the comparison column is replaced by the ref_group and test_group columns.p_val column for raw p-values, a p_adj column for adjusted p-values, and a log2fc column for log2 fold change, if applicable. This consistence in column naming reduces the cognitive load.get_tidy_result() and get_raw_result() to get the tidy result tibble and the raw result list from a glystats result object. These functions are useful to be used in pipes.filter_sig_vars() to filter the experiment using the results from glystats DEA functions to keep only significant variables.gly_fold_change() now supports multiple groups.gly_anova() and gly_kruskal() now have a log2fc column for log2 fold change.ref_group and test_group columns in the tidy_result tibble of gly_limma() even for 2 groups now.gly_kruskal() failed to work with experiments with 2 groups.gly_anova() and gly_kruskal() have NAs in the post_hoc_test tibble.glyrepr to dependencies to fix the result printing bug.could not find function "%||%" bug in gly_anova().gly_anova(), gly_kruskal(), gly_ttest(), gly_wilcox(), gly_cox(), gly_roc(), and their gly_xxx_() counterparts are now more robust. Previously, if the model failed to be fitted for any variable, the functions stopped with an error. Now, they only issue a warning and continue, assigning NAs in the results for those failed variables.gly_plsda() now uses the ropls package as its backend instead of mixOmics. This changes the class of the object returned in the raw_result list element, which may affect downstream code that uses the raw result directly.gly_oplsda() via the new perm_test parameter to assess model significance.gly_oplsda() now includes a pcorr column containing p-values for the correlation coefficients of features with the predictive component.gly_oplsda() and gly_plsda(), increasing flexibility for smaller datasets.gly_limma() that could cause duplicated messages to be printed.gly_oplsda() where the column for the first orthogonal component score (o1) could be missing from the result.gly_enrich_go() for a cleaner console output.glystats S3 classes from the tibbles in the tidy_result, simplifying the output object structure.This version introduces new API for all functions.
Briefly, the return_raw parameter is removed,
and all functions now return a list with two elements:
tidy_result and raw_result.
The concrete types of tidy_result and raw_result depend.
tidy_result can be a tibble, or a list of tibbles.
raw_result can be a single object returned by the underlying statistical functions,
or a list of such objects.
This update makes glystats easier to use.
And more importantly, it allows the glyvis package to access the raw results directly.
tidy_result.gly_xxx_() functions now accept a character vector as the groups parameter.gly_consensus_clustering() sends plots to the plot panel when output_file is NULL.
This is an inconsistent behavior compared to other functions, and it has been fixed.gly_consensus_clustering() to emphasize the importance of output_file.gly_umap() and gly_umap_() consistent.gly_tsne() and gly_tsne_().gly_anova(), gly_kruskal(), gly_limma(), gly_ttest(), and gly_wilcox().gly_anova() and gly_kruskal() now return lists of two tibbles, one for main test and the other for post-hoc test.method, dist_method parameters from gly_hclust().max_iter, theta parameters from gly_tsne().n_epochs, learning_rate, metric parameters from gly_umap().gly_xxx() functions now have a lower API counterpart gly_xxx_()
that works with matrices directly, providing more flexibility for users who don't use the glyexp package.gly_limma() and gly_limma_() to perform differential analysis using the limma package.gly_kmeans() and gly_kmeans_() to perform K-means clustering.gly_wgcna() and gly_wgcna_() to perform WGCNA analysis.gly_consensus_clustering() and gly_consensus_clustering_() to perform consensus clustering.gly_cox() and gly_cox_() to fit Cox proportional hazards model for survival analysis.ref_group parameter to gly_ttext() and gly_wilcox().parameters package.