minor additions and bugfixes

DESCRIPTION

@@ -1,6 +1,6 @@
 Package: BayesSpace
-Version: 1.5.0
-Date: 2021-08-05
+Version: 1.5.1
+Date: 2021-11-03
 Title: Clustering and Resolution Enhancement of Spatial Transcriptomes
 Authors@R: c(
     person("Edward", "Zhao", role=c("aut"), email="[email protected]"),

NEWS.md

@@ -1,3 +1,23 @@
+# BayesSpace 1.5.1
+
+## Minor improvements and fixes
+
+- Update `readVisium` to work properly when feature names include whitespaces
+- Fix rounding of w_i sum in MCMC
+- Add details to vignette related to `spatialEnhance`
+
+# BayesSpace 1.5.0
+
+## New Bioconductor devel (3.15)
+
+- Version numbering change with Bioconductor version bump
+
+# BayesSpace 1.4.0
+
+## New Bioconductor release (3.14)
+
+- Version numbering change with Bioconductor version bump
+
 # BayesSpace 1.3.1
 
 ## Minor improvements and fixes
@@ -13,9 +33,9 @@
 
 # BayesSpace 1.2.0
 
-## Minor improvements and fixes
+## New Bioconductor release (3.12)
 
-- Stable Bioconductor release 3.12
+- Version numbering change with Bioconductor version bump
 
 # BayesSpace 1.1.2
 

R/readVisium.R

@@ -52,13 +52,13 @@ readVisium <- function(dirname) {
     rownames(colData) <- colData$spot
     colData <- colData[colData$in_tissue > 0, ]
 
-    rowData <- read.table(file.path(matrix_dir, "features.tsv.gz"), header=FALSE)
+    rowData <- read.table(file.path(matrix_dir, "features.tsv.gz"), header=FALSE, sep = "\t")
     colnames(rowData) <- c("gene_id", "gene_name", "feature_type")
     rowData <- rowData[, c("gene_id", "gene_name")]
     rownames(rowData) <- scater::uniquifyFeatureNames(rowData$gene_id, rowData$gene_name)
 
     counts <- Matrix::readMM(file.path(matrix_dir, "matrix.mtx.gz"))
-    barcodes <- read.table(file.path(matrix_dir, "barcodes.tsv.gz"), header=FALSE)
+    barcodes <- read.table(file.path(matrix_dir, "barcodes.tsv.gz"), header=FALSE, sep = "\t")
     colnames(counts) <- barcodes$V1
     rownames(counts) <- rownames(rowData)
     counts <- counts[, rownames(colData)]

src/RcppExports.cpp

@@ -6,6 +6,11 @@
 
 using namespace Rcpp;
 
+#ifdef RCPP_USE_GLOBAL_ROSTREAM
+Rcpp::Rostream<true>&  Rcpp::Rcout = Rcpp::Rcpp_cout_get();
+Rcpp::Rostream<false>& Rcpp::Rcerr = Rcpp::Rcpp_cerr_get();
+#endif
+
 // iterate
 List iterate(arma::mat Y, List df_j, int nrep, int n, int d, double gamma, int q, arma::vec init, NumericVector mu0, arma::mat lambda0, double alpha, double beta);
 RcppExport SEXP _BayesSpace_iterate(SEXP YSEXP, SEXP df_jSEXP, SEXP nrepSEXP, SEXP nSEXP, SEXP dSEXP, SEXP gammaSEXP, SEXP qSEXP, SEXP initSEXP, SEXP mu0SEXP, SEXP lambda0SEXP, SEXP alphaSEXP, SEXP betaSEXP) {

src/cluster.cpp

@@ -261,7 +261,7 @@ List iterate_t(arma::mat Y, List df_j, int nrep, int n, int d, double gamma, int
     NumericVector Ysums;
     for (int k = 1; k <= q; k++){
       uvec index_1k = find(df_sim_z.row(i-1) == k);
-      int n_i = sum(w(index_1k)); 
+      double n_i = sum(w(index_1k)); 
       mat Yrows = Y.rows(index_1k);
       Yrows.each_col() %= w(index_1k);
       Ysums = sum(Yrows, 0);
@@ -359,7 +359,7 @@ List iterate_t_vvv(arma::mat Y, List df_j, int nrep, int n, int d, double gamma,
     NumericVector Ysums;
     for (int k = 1; k <= q; k++){
       uvec index_1k = find(df_sim_z.row(i-1) == k);
-      int n_i = sum(w(index_1k)); 
+      double n_i = sum(w(index_1k)); 
       mat Yrows = Y.rows(index_1k);
       Yrows.each_col() %= w(index_1k);
       Ysums = sum(Yrows, 0);
@@ -489,7 +489,7 @@ List iterate_deconv(arma::mat Y, List df_j, bool tdist, int nrep, int n, int n0,
     NumericVector Ysums;
     for (int k = 1; k <= q; k++){
       uvec index_1k = find(z == k);
-      int n_i = sum(w(index_1k)); 
+      double n_i = sum(w(index_1k)); 
       mat Yrows = Y.rows(index_1k);
       Yrows.each_col() %= w(index_1k);
       Ysums = sum(Yrows, 0);

vignettes/BayesSpace.Rmd

@@ -168,6 +168,13 @@ components, and add these PCs as well as predicted cluster labels at subspot
 resolution to a new `SingleCellExperiment`. As with our demonstration of
 `spatialCluster()` above, we are using fewer iterations for the purpose of this
 example (`nrep=1000`) than we recommend in practice (`nrep=100000` or greater).
+Note that the `jitter_scale` parameter should be tuned so that proposals for 
+updating subspot-level expression are accepted around 30% of the time. This can
+be evaluated using `mcmcChain(melanoma.enhanced, "Ychange")`, where the chain 
+should stabilize to 0.25-0.40. Typically 1000-2500 iterations are sufficient to 
+evaluate if `jitter_scale` should be increased if acceptance is too high or 
+decreased if acceptance is too low. After tuning, proceed to a full run of 
+`spatialEnhance` with more iterations.
 
 ```{r enhance, eval=TRUE}
 melanoma.enhanced <- spatialEnhance(melanoma, q=4, platform="ST", d=7,