BayesSpace v1.11.3

.Rbuildignore

@@ -1,3 +1,5 @@
+^renv$
+^renv\.lock$
 ^.*\.Rproj$
 ^\.Rproj\.user$
 ^inst/profiling$

.gitignore

@@ -61,3 +61,5 @@ docs/
 
 *.nb.html
 *.rds
+
+src/indicators

DESCRIPTION

@@ -1,6 +1,6 @@
 Package: BayesSpace
-Version: 1.11.0
-Date: 2023-05-03
+Version: 1.13.6
+Date: 2024-04-02
 Title: Clustering and Resolution Enhancement of Spatial Transcriptomes
 Authors@R: c(
     person("Edward", "Zhao", role=c("aut"), email="[email protected]"),
@@ -33,20 +33,23 @@ Imports:
     DirichletReg,
     xgboost,
     utils,
-    tidyverse,
+    dplyr,
+    ggplot2,
+    tibble,
+    tidyr,
     scales,
     BiocFileCache,
     BiocSingular,
     BiocParallel
 License: MIT + file LICENSE
-RoxygenNote: 7.2.3
+RoxygenNote: 7.3.2
 LinkingTo: 
     Rcpp,
     RcppArmadillo,
     RcppDist,
     RcppProgress
 NeedsCompilation: yes
-SystemRequirements: C++11
+SystemRequirements: C++17
 Encoding: UTF-8
 Suggests: 
     testthat,

NAMESPACE

@@ -1,10 +1,15 @@
 # Generated by roxygen2: do not edit by hand
 
+export(adjustClusterLabels)
 export(clusterPlot)
+export(computeCorr)
+export(coreTune)
+export(counts2h5)
 export(enhanceFeatures)
 export(exampleSCE)
 export(featurePlot)
 export(getRDS)
+export(mapSubspot2Ref)
 export(mcmcChain)
 export(paraLapply)
 export(qPlot)
@@ -22,6 +27,9 @@ importFrom(BiocParallel,SerialParam)
 importFrom(BiocParallel,SnowParam)
 importFrom(BiocParallel,bplapply)
 importFrom(BiocSingular,ExactParam)
+importFrom(BiocSingular,FastAutoParam)
+importFrom(BiocSingular,IrlbaParam)
+importFrom(BiocSingular,RandomParam)
 importFrom(DirichletReg,DR_data)
 importFrom(DirichletReg,DirichReg)
 importFrom(Matrix,colSums)
@@ -48,9 +56,16 @@ importFrom(SummarizedExperiment,assay)
 importFrom(SummarizedExperiment,assayNames)
 importFrom(SummarizedExperiment,colData)
 importFrom(SummarizedExperiment,rowData)
+importFrom(arrow,read_parquet)
 importFrom(assertthat,assert_that)
 importFrom(coda,mcmc)
+importFrom(dplyr,"%>%")
+importFrom(dplyr,case_when)
+importFrom(dplyr,group_by)
 importFrom(dplyr,inner_join)
+importFrom(dplyr,mutate)
+importFrom(dplyr,n)
+importFrom(dplyr,select)
 importFrom(ggplot2,aes_)
 importFrom(ggplot2,coord_equal)
 importFrom(ggplot2,geom_line)
@@ -65,18 +80,23 @@ importFrom(ggplot2,theme_bw)
 importFrom(ggplot2,theme_void)
 importFrom(ggplot2,xlab)
 importFrom(ggplot2,ylab)
+importFrom(magrittr,"%>%")
 importFrom(mclust,Mclust)
 importFrom(mclust,mclustBIC)
+importFrom(microbenchmark,microbenchmark)
+importFrom(parallel,detectCores)
 importFrom(purrr,compact)
 importFrom(purrr,discard)
 importFrom(purrr,imap)
 importFrom(purrr,keep)
 importFrom(purrr,map)
 importFrom(rhdf5,h5createDataset)
 importFrom(rhdf5,h5createFile)
+importFrom(rhdf5,h5createGroup)
 importFrom(rhdf5,h5ls)
 importFrom(rhdf5,h5read)
 importFrom(rhdf5,h5write)
+importFrom(rjson,fromJSON)
 importFrom(scales,muted)
 importFrom(scater,logNormCounts)
 importFrom(scater,runPCA)
@@ -94,6 +114,7 @@ importFrom(stats,rnbinom)
 importFrom(stats,rnorm)
 importFrom(stats,runif)
 importFrom(tibble,as_tibble)
+importFrom(tibble,column_to_rownames)
 importFrom(tidyr,uncount)
 importFrom(utils,download.file)
 importFrom(utils,read.csv)

NEWS.md

@@ -1,3 +1,21 @@
+# BayesSpace 1.15.0
+
+## New Bioconductor release (3.20)
+
+### Major updates
+
+- Accelerate resolution enhancement with multithreaded `spatialEnhance`
+- Improve mixing of MCMC for `spatialEnhance` with adaptive MCMC
+- Support VisiumHD
+
+### Minor improvements and fixes
+
+- Support SpaceRanger v2.0+
+- Find the optimal number of cores with `coreTune` before enhancing resolution
+- Adjust the proportion of samples to be removed during burnin using `adjustClusterLabels`
+- Faster neighbor finding for subspots
+- Customize resolution enhancement for VisiumHD data
+
 # BayesSpace 1.5.1
 
 ## Minor improvements and fixes
@@ -81,10 +99,10 @@
 ## Minor improvements and fixes
 
 - Addressed reviewer concerns (https://github.com/Bioconductor/Contributions/issues/1624)
-    * Updated stop/warning/message statements to remove redundancies and
-      unnecessary use of `paste()`.
-    * Removed inline conditional statements.
-    * Cache downloaded RDS in `getRDS()` (see above).
+  - Updated stop/warning/message statements to remove redundancies and
+    unnecessary use of `paste()`.
+  - Removed inline conditional statements.
+  - Cache downloaded RDS in `getRDS()` (see above).
 - `spatialCluster()` and `spatialEnhance()` handle the edge case where only one
   iteration is kept after excluding burn-in.
 - The `coda::mcmc` object returned by `mcmcChain()` now specifies the thinning

R/BayesSpace.R

@@ -1,26 +1,25 @@
 #' BayesSpace: A package for processing spatial transcriptomes
-#' 
+#'
 #' Tools for clustering and enhancing the resolution of spatial gene expression
 #' experiments. BayesSpace clusters a low-dimensional representation of the gene
 #' expression matrix, incorporating a spatial prior to encourage neighboring
 #' spots to cluster together. The method can enhance the resolution of the
 #' low-dimensional representation into "sub-spots", for which features such as
 #' gene expression or cell type composition can be imputed.
-#' 
+#'
 #' @details
 #' For an overview of the functionality provided by the package, please see the
 #' vignette:
 #' \code{vignette("BayesSpace", package="BayesSpace")}
 #'
-#' 
-#' @docType package
+#'
 #' @name BayesSpace
-#' 
+#'
 #' @keywords internal
 #'
 #' @importFrom Rcpp evalCpp
 #' @importFrom SummarizedExperiment assayNames
 #' @importFrom SingleCellExperiment altExpNames
 #' @useDynLib BayesSpace
 #' @useDynLib BayesSpace, .registration=TRUE
-NULL
+"_PACKAGE"

R/RcppExports.R

@@ -1,23 +1,31 @@
 # Generated by using Rcpp::compileAttributes() -> do not edit by hand
 # Generator token: 10BE3573-1514-4C36-9D1C-5A225CD40393
 
-iterate <- function(Y, df_j, nrep, n, d, gamma, q, init, mu0, lambda0, alpha, beta) {
-    .Call(`_BayesSpace_iterate`, Y, df_j, nrep, n, d, gamma, q, init, mu0, lambda0, alpha, beta)
+iterate <- function(Y, df_j, nrep, thin, n, d, gamma, q, init, mu0, lambda0, alpha, beta) {
+    .Call(`_BayesSpace_iterate`, Y, df_j, nrep, thin, n, d, gamma, q, init, mu0, lambda0, alpha, beta)
 }
 
-iterate_vvv <- function(Y, df_j, nrep, n, d, gamma, q, init, mu0, lambda0, alpha, beta) {
-    .Call(`_BayesSpace_iterate_vvv`, Y, df_j, nrep, n, d, gamma, q, init, mu0, lambda0, alpha, beta)
+iterate_vvv <- function(Y, df_j, nrep, thin, n, d, gamma, q, init, mu0, lambda0, alpha, beta) {
+    .Call(`_BayesSpace_iterate_vvv`, Y, df_j, nrep, thin, n, d, gamma, q, init, mu0, lambda0, alpha, beta)
 }
 
-iterate_t <- function(Y, df_j, nrep, n, d, gamma, q, init, mu0, lambda0, alpha, beta) {
-    .Call(`_BayesSpace_iterate_t`, Y, df_j, nrep, n, d, gamma, q, init, mu0, lambda0, alpha, beta)
+iterate_t <- function(Y, df_j, nrep, thin, n, d, gamma, q, init, mu0, lambda0, alpha, beta) {
+    .Call(`_BayesSpace_iterate_t`, Y, df_j, nrep, thin, n, d, gamma, q, init, mu0, lambda0, alpha, beta)
 }
 
-iterate_t_vvv <- function(Y, df_j, nrep, n, d, gamma, q, init, mu0, lambda0, alpha, beta) {
-    .Call(`_BayesSpace_iterate_t_vvv`, Y, df_j, nrep, n, d, gamma, q, init, mu0, lambda0, alpha, beta)
+iterate_t_vvv <- function(Y, df_j, nrep, thin, n, d, gamma, q, init, mu0, lambda0, alpha, beta) {
+    .Call(`_BayesSpace_iterate_t_vvv`, Y, df_j, nrep, thin, n, d, gamma, q, init, mu0, lambda0, alpha, beta)
 }
 
-iterate_deconv <- function(Y, df_j, tdist, nrep, n, n0, d, gamma, q, init, subspots, verbose, jitter_scale, c, mu0, lambda0, alpha, beta, thread_num = 1L) {
-    .Call(`_BayesSpace_iterate_deconv`, Y, df_j, tdist, nrep, n, n0, d, gamma, q, init, subspots, verbose, jitter_scale, c, mu0, lambda0, alpha, beta, thread_num)
+iterate_deconv <- function(subspot_positions, dist, spot_neighbors, Y, tdist, nrep, thin, n, n0, d, gamma, q, init, subspots, verbose, jitter_scale, adapt_before, c, mu0, lambda0, alpha, beta, thread_num = 1L) {
+    .Call(`_BayesSpace_iterate_deconv`, subspot_positions, dist, spot_neighbors, Y, tdist, nrep, thin, n, n0, d, gamma, q, init, subspots, verbose, jitter_scale, adapt_before, c, mu0, lambda0, alpha, beta, thread_num)
+}
+
+map_subspot2ref <- function(subspot_coords, ref_coords, thread_num = 1L) {
+    .Call(`_BayesSpace_map_subspot2ref`, subspot_coords, ref_coords, thread_num)
+}
+
+compute_corr <- function(m1, m2, thread_num = 1L) {
+    .Call(`_BayesSpace_compute_corr`, m1, m2, thread_num)
 }
 

R/mcmcChain.R

@@ -111,7 +111,12 @@ NULL
 #' @importFrom purrr map
 .read_chain <- function(h5.fname, params = NULL, is.enhanced = FALSE) {
     if (is.null(params)) {
+      if (is.enhanced)
+        params <- c("z", "mu", "lambda", "weights", "Y")
+      else {
         params <- h5ls(h5.fname)$name
+        params <- params[params != "plogLik"]
+      }
     }
 
     .read_param <- function(par.name) {
@@ -178,16 +183,6 @@ NULL
 .clean_chain <- function(out, method = c("cluster", "enhance"), thin=100) 
 {
     method <- match.arg(method)
-    n_iter <- nrow(out$z)  # this is technically n_iters / 100 for enhance
-
-    ## Only one iteration included; need to cast vectors back to matrices
-    if (is.null(n_iter)) {
-        out$z <- matrix(out$z, nrow=1)
-        out$mu <- matrix(out$mu, nrow=1)
-        if ("weights" %in% names(out)) {
-            out$weights <- matrix(out$mu, nrow=1)
-        }
-    }
 
     n <- ncol(out$z)
     d <- ncol(out$lambda[[1]])
@@ -206,22 +201,9 @@ NULL
     colnames(out$plogLik) <- c("pLogLikelihood")
     if (method == "enhance") {
         out$Y <- .flatten_matrix_list(out$Y, "Y", n, d)
-        out$Ychange <- as.matrix(out$Ychange)
-        colnames(out$Ychange) <- c("Ychange")
-    }
-
-    if (method == "enhance") {
-        ## manually thin mu until updated in c++; 
-        ## keep init values for consistency with others
-        thinned_idx <- c(1, seq(thin, (n_iter - 1) * thin, thin))
-        out$mu <- out$mu[thinned_idx, ]
 
-        ## Subset of a one-column matrix is a vector, not a matrix
-        out$Ychange <- as.matrix(out$Ychange[thinned_idx, ])
+        out$Ychange <- as.matrix(out$Ychange)
         colnames(out$Ychange) <- c("Ychange")
-
-        out$plogLik <- as.matrix(out$plogLik[thinned_idx, ])
-        colnames(out$plogLik) <- c("pLogLikelihood")
     }
 
     out

R/parallelize.R

@@ -107,7 +107,7 @@ paraLapply <- function(X, FUN, BPPARAM = NULL, cores = 1L, type = c("serial", "f
   if (verbose) {
     message(sprintf("[DEBUG] Provided effective arguments for the customized function: %s", .list2vec(.extra.args)))
   }
-
+  
   do.call(
     bplapply,
     c(

R/qTune.R

@@ -86,7 +86,7 @@ qTune <- function(sce, qs = seq(3, 7), burn.in = 100, nrep = 1000, cores = 1L, .
 
   ## Get neighbors
   platform <- ifelse(is.null(args$platform), "Visium", args$platform)
-  df_j <- .find_neighbors(sce, platform)
+  df_j <- .find_neighbors(sce, platform)[[2]]
 
   ## Parse args from ... for cluster initialization
   init.args <- c("init", "init.method")
@@ -115,7 +115,7 @@ qTune <- function(sce, qs = seq(3, 7), burn.in = 100, nrep = 1000, cores = 1L, .
     },
     cores = cores,
     type = "fork",
-    verbose = FALSE
+    verbose = TRUE
   )
 
   attr(sce, "q.logliks") <- do.call(rbind, logliks)