Fix prepare_ funcs. Add run_integration

DESCRIPTION

@@ -81,15 +81,18 @@ Suggests:
     downloadR,
     HPOExplorer,
     Nebulosa,
-    arrow
+    arrow,
+    ggsankey,
+    Hmisc
 Remotes:
     github::neurogenomics/scNLP,
     github::neurogenomics/scKirby,
     github::neurogenomics/KGExplorer,
     github::neurogenomics/HPOExplorer,
     github::RajLabMSSM/echotabix,
     github::RajLabMSSM/downloadR,
-    github::satijalab/seurat-wrappers
+    github::satijalab/seurat-wrappers,
+    github::davidsjoberg/ggsankey
 RoxygenNote: 7.3.1
 VignetteBuilder: knitr 
 Config/testthat/edition: 3

NAMESPACE

@@ -1,6 +1,7 @@
 # Generated by roxygen2: do not edit by hand
 
 export(add_cluster_colors)
+export(add_mixingmetric)
 export(adjust_zstat)
 export(dt_to_granges)
 export(find_neighbors)
@@ -15,11 +16,13 @@ export(get_mhc_genes)
 export(get_top_factors)
 export(get_top_features)
 export(get_top_traits)
+export(get_variance_explained)
 export(import_abc)
 export(is_granges)
 export(iterate_gsea)
 export(iterate_lm)
 export(magma_matrix)
+export(map_id_sep)
 export(map_phenotypes)
 export(map_snps2genes)
 export(opengwas_meta)
@@ -28,10 +31,12 @@ export(phenomix_query)
 export(phenomix_query_batched)
 export(plot_dag_enrich)
 export(plot_enrichment)
+export(plot_factors_sankey)
 export(plot_feature_density)
 export(plot_highlights)
 export(plot_hpo_pseudotime)
 export(plot_ontological_velocity)
+export(plot_preservation_histo)
 export(plot_reduction)
 export(plot_top_celltypes)
 export(plot_trait_cor)
@@ -44,12 +49,14 @@ export(run_autoencoder)
 export(run_dag_enrich)
 export(run_gprofiler)
 export(run_imputation)
+export(run_integration)
 export(run_lda)
 export(run_mofa2)
 export(run_pca)
 export(run_sparsesvd)
 export(run_umap)
 export(run_variancePartition)
+export(scale_color_nightlight)
 export(theme_nightlight)
 import(GenomicFiles)
 import(HPOExplorer)

R/add_mixingmetric.R

@@ -0,0 +1,41 @@
+#' Add mixing metric
+#' 
+#' Compute the mixing metric for a given reduction and add it to the metadata 
+#' of a \link{Seurat} object. See \link[Seurat]{MixingMetric} for more details.
+#' @param invert Invert the mixing metric such that higher values 
+#' indicate better mixing.
+#' @param normalise Normalise the mixing metric to the range [0,1].
+#' @inheritParams Seurat::MixingMetric
+#' @inheritDotParams Seurat::MixingMetric
+#' @export
+add_mixingmetric <- function(obj,
+                             reduction,
+                             new_col=paste0("MixingMetric_",reduction),
+                             grouping.var = "source",
+                             k = 5,
+                             max.k = 300, 
+                             dims = NULL,
+                             invert = TRUE,
+                             normalise = FALSE,
+                             ...){
+    if(is.null(dims)){
+        dims <- seq(ncol(obj[[reduction]]))
+    }
+    message("Computing mixing metric for reduction: ",reduction,
+            " (",max(dims)," dimensions)")
+    obj@meta.data[[new_col]] <- Seurat::MixingMetric(
+        obj,
+        reduction = reduction,
+        grouping.var = grouping.var,
+        k = k,
+        max.k = max.k, 
+        dims = dims,
+        ...) 
+    if(isTRUE(invert)){
+        obj@meta.data[[new_col]] <- max.k - obj@meta.data[[new_col]]
+    }
+    if(isTRUE(normalise)){
+        obj@meta.data[[new_col]] <- obj@meta.data[[new_col]] / max.k
+    } 
+    return(obj)
+}

R/color_nightlight.R

@@ -1,4 +1,31 @@
-color_nightlight <- function(n = 40){
-    ggplot2::scale_color_gradientn(
-        colors=pals::gnuplot(n = n)[-(seq(as.integer(n*0.2)))])
+#' Scale color nightlight
+#' 
+#' Add color gradient with nightlight colors.
+#' @export
+#' @param reverse Reverse the color gradient.
+#' @param alpha Transparency of the colors.
+#' @inheritParams pals::gnuplot
+#' @inheritParams ggplot2::scale_color_gradientn
+#' @inheritDotParams ggplot2::scale_color_gradientn
+scale_color_nightlight <- function(n = 3,
+                                   colors=pals::gnuplot(n = n+1)[
+                                       -(seq(as.integer((n+1)*0.2)))
+                                       ],
+                                   reverse=FALSE,
+                                   alpha=NULL,
+                                   ...){
+    if(!is.null(alpha)) {
+        if(length(alpha)==1){
+            colors <- ggplot2::alpha(colors,alpha)    
+        } else if(length(alpha)==length(colors)){
+            for(i in seq_along(colors)){
+                colors[[i]] <- ggplot2::alpha(colors[[i]],alpha[i])
+            }
+        } else {
+            stopper("Length of alpha must be 1 or equal to length of colors.")
+        }
+    }
+    if(isTRUE(reverse)) colors <- rev(colors)
+    ggplot2::scale_color_gradientn(colors=colors,
+                                   ...)
 }

R/get_correlated_factors.R

@@ -0,0 +1,21 @@
+get_correlated_factors <- function(obj,
+                                   keys,
+                                   metadata_var,
+                                   p_threshold=.05,
+                                   r_threshold=.1){
+    requireNamespace("Hmisc")
+    obsm <- scKirby::get_obsm(obj,
+                              keys = keys,
+                              n=1)
+    obsm <- cbind(obsm, 
+                  nFeature_score=obj$nFeature_score[rownames(obsm)])
+    obsm_rcor <- Hmisc::rcorr(obsm)
+    obsm_cor_sig <- sort(
+        abs(obsm_rcor$r[metadata_var,][obsm_rcor$P[metadata_var,]<p_threshold]), 
+        decreasing = TRUE
+    ) 
+    omit_dims <- as.numeric(gsub(".*_","",
+                                 names(obsm_cor_sig[obsm_cor_sig>r_threshold]))
+                            )
+    return(omit_dims)
+}

R/get_variance_explained.R

@@ -0,0 +1,40 @@
+#' Get variance explained
+#' 
+#' Get the proportion of total variance explained
+#'  by a given dimensionality reduction 
+#'  (0-1 where 1 indicates 100% of variance explained). 
+#' @export
+#' @examples
+#' obj <- get_HPO()
+#' out <- get_variance_explained(obj)
+get_variance_explained <- function(obj,
+                                   reduction = names(obj@reductions)[1],
+                                   layer = "scale.data",
+                                   dims=NULL
+                                   ){
+    dr <- obj[[reduction]] 
+    mat <- Seurat::GetAssayData(obj, layer = layer)
+    total_variance <- sum(matrixStats::rowVars(mat)) 
+    ## EigenValues
+    if(length(dr@stdev)==0){
+        messager("Computing stdev for reduction.")
+        dr@stdev <- apply(obj[[reduction]]@cell.embeddings,2,
+                          stats::sd)|>unname()
+    }
+    eigValues <- (dr@stdev)^2|>`names<-`(colnames(dr@cell.embeddings)) 
+    if(!is.null(dims)) {
+        if(length(dims)>length(eigValues) || 
+           max(dims)>length(eigValues) ){
+            stopper(
+                "The length of dims must be less than or equal to",
+                "the number of dimensions."
+                )
+        }  
+        eigValues <- eigValues[dims]
+    }
+    varExplained <- eigValues / total_variance
+    messager("Proportion of total variance explained by",
+             length(eigValues),"dimensions:",
+             round(sum(varExplained),3))
+    return(varExplained)
+}

R/map_xref.R

@@ -12,7 +12,8 @@ map_xref <- function(dat,
             if(length(r)==0) NA else unlist(r)
         })]
     }
+    dat[,(new_col):=map_id_sep(get(new_col))]
     dat[get(new_col)=="NA",(new_col):=NA]
     messager("% of non-NA rows:",
              round(sum(!is.na(dat[[new_col]]))/nrow(dat)*100,2),v=verbose)
-}
+}