jgarces02 / flowct Goto Github PK
View Code? Open in Web Editor NEWFlowCT package
FlowCT package
Add (clinical, FISH, treatment, % cells...) metadata for doing clustering... it's expected to make better clusters.
processx is a dependency, it should be added...Thanks!
In my windows 10 PC I had to add the following line just before downloading
options(download.file.method = "libcurl")
Not sure if always required
multidensity <- function(fcs.SCE, assay.i, show.markers = "all", color.by = NULL, subsampling = NULL, interactive = F, ridgeline.lim = 15, colors = NULL){
if(show.markers == "all") show.markers <- rownames(fcs.SCE)
if(!is.null(subsampling)) suppressMessages(fcs.SCE <- sub.samples(fcs.SCE, subsampling = subsampling))
data <- t(assay(fcs.SCE, i = assay.i))
data2 <- cbind(data, colData(fcs.SCE))
# prepare tables: for plotting and with median values for each marker
median_df <- data.frame(antigen = show.markers, median = apply(data[,show.markers], 2, median))
ggdf <- as.data.frame(melt(as.data.table(data2), measure.vars = show.markers, value.name = "expression", variable.name = "antigen"))
if(is.null(colors)) colors <- div.colors(unique(length(ggdf[,color.by])))
if(length(unique(fcs.SCE$filename)) > ridgeline.lim){
g <- ggplot(data = ggdf[grepl(paste0(show.markers, collapse = "|"), ggdf$antigen),],
aes_string(x = "expression", color = color.by, group = "filename")) +
# geom_density(size = 0.5) +
stat_density(geom = "line", position = "identity", size = 0.5) +
facet_wrap(~ antigen, scales = "free") +
geom_vline(data = median_df, aes(xintercept = median), linetype = 2, color = "gray55") +
scale_color_manual(name = color.by, values = colors) +
theme_minimal() + theme(axis.text.x = element_text(angle = 90, hjust = 1),
strip.text = element_text(size = 7), axis.text = element_text(size = 5))
if(interactive) ggplotly(g) else print(g)
}else{
suppressMessages(print(ggplot(ggdf, aes_string(x = "expression", y = "filename")) +
geom_density_ridges(alpha = 0.7) +
facet_wrap(~ antigen, scales = "free") +
geom_vline(data = median_df, aes(xintercept = median), linetype = 2, color = "gray55") +
theme_minimal() + theme(axis.text.x = element_text(angle = 90, hjust = 1),
strip.text = element_text(size = 7), axis.text = element_text(size = 7))))
}
}
Hi, some issues in this part.
Data should be scaled to ensure correct clustering; I added set_seed to ensure consistent results, I removed the first if loop due to errors. Now everything seems fine and working.
fsom.clustering <- function(fcs.SE, assay.i = "normalized", scale.data= TRUE, markers.to.use = rownames(fcs.SE), markers.to.plot = NULL, k.metaclustering = 40, metaclustering.name = NULL){
require(FlowSOM)
set.seed(1234)
data <- as.flowSet.SE(fcs.SE, assay.i)
## FSOM clustering
cat("Calculating SOM clustering...\n")
fsom <- suppressMessages(ReadInput(data, transform = FALSE, scale = scale.data)) #read data
fsom <- suppressMessages(BuildSOM(fsom, colsToUse = markers.to.use)) #build SOM
cat("Building MST...\n")
fsom <- suppressMessages(BuildMST(fsom, tSNE = TRUE, silent = T)) #build MST for visualization of clustering
#plot the MST to evaluate the marker fluorescence (or general tree) intensity for each SOM
if(!is.null(markers.to.plot)){
if(markers.to.plot == "tree"){
PlotStars(fsom)
}else{
for(marker in markers.to.plot) PlotMarker(fsom, marker)
}
}
## Metaclustering
if(!is.null(k.metaclustering)){
cat("Metaclustering...\n")
if(!is.null(metaclustering.name)){
mc <- suppressMessages(ConsensusClusterPlus::ConsensusClusterPlus(t(fsom$map$codes), maxK = k.metaclustering, reps = 100,
pItem = 0.9, pFeature = 1, title = metaclustering.name, plot = "pdf",
clusterAlg = "hc", innerLinkage = "average", finalLinkage = "average",
distance = "euclidean", seed = 333, verbose = F))
}else{
mc <- suppressMessages(ConsensusClusterPlus::ConsensusClusterPlus(t(fsom$map$codes), maxK = k.metaclustering, reps = 100,
pItem = 0.9, pFeature = 1, title = "consensus_plots", plot = "pdf",
clusterAlg = "hc", innerLinkage = "average", finalLinkage = "average",
distance = "euclidean", seed = 333, verbose = F))
unlink("consensus_plots", recursive = TRUE)
}
#get cluster ids for each cell
code_clustering1 <- mc[[k.metaclustering]]$consensusClass %>% as.factor()
cell_clustering1 <- code_clustering1[fsom$map$mapping[,1]]
#add clustering to original MST and color by cluster colors
PlotStars(fsom, backgroundValues = code_clustering1, backgroundColor = alpha(div.colors(length(code_clustering1)), alpha = 0.7))
return(list(fsom = fsom, metaclusters = cell_clustering1, plotStars_value = code_clustering1))
}else{
return(fsom)
}
}
new_names = NULL
doesn't show new names.Error in `colnames<-`(`*tmp*`, value = new_names) :
attempt to set 'colnames' on an object with less than two dimensions
expr_no_transfL <- expr_no_transf[metadata_sc$FlowSOM == "CD8p",]
MATCH...
From flowStats ---> gaussNorm (already implemented) and wrapSet: both based on per-channel landarmks.
From iFlow ---> gpaSet: multidimensional normalization method using the generalized Procrustes analysis.
This part should be replaced by the specific function after preparing the original SCE:
#to be replaced
fcsL <- fcs1000[,fcs1000$SOM_named == "lymphocytes"]
metadata(fcsL)$subclustering <- "lymphocytes"
###new code to be included in the script
fcs$SOM_named <- clusters.rename(fcs$SOM, cluster = replacedata$original_cluster, name = replacedata$new_cluster)
fcsL<-subset(fcs, ,SOM_named=="lymphocytes")
> fcs <- normalization.flw(fcs.SCE = fcs, marker.to.norm = c("CCR6", "CCR4"),
+ norm.method = "harmony", var.to.use = "patient_id")
error: Mat::init(): requested size is too large
*** caught segfault ***
address (nil), cause 'memory not mapped'
Traceback:
1: .External(list(name = "CppMethod__invoke_void", address = <pointer: 0x117fd830>, dll = list(name = "Rcpp", path = "/home/cdasilvam/FlowCT.v2_environment1/renv/library/R-4.0/x86_64-pc-linux-gnu/Rcpp/libs/Rcpp.so", dynamicLookup = TRUE, handle = <pointer: 0x1b6c8e90>, info = <pointer: 0x19cfe20>), numParameters = -1L), <pointer: 0x29d889b0>, <pointer: 0x361e00c0>, .pointer, ...)
2: harmonyObj$setup(data_mat, phi, phi_moe, Pr_b, sigma, theta, max.iter.cluster, epsilon.cluster, epsilon.harmony, nclust, tau, block.size, lambda_mat, verbose)
3: HarmonyMatrix(data_mat = assay(fcs.SCE, assay.i), meta_data = colData(fcs.SCE), vars_use = var.to.use, do_pca = F, verbose = F)
4: normalization.flw(fcs.SCE = fcs, marker.to.norm = c("CCR6", "CCR4"), norm.method = "harmony", var.to.use = "patient_id")
Possible actions:
1: abort (with core dump, if enabled)
2: normal R exit
3: exit R without saving workspace
4: exit R saving workspace
When creating it, use surface_markers
instead select specific metadata cols.
If filenames has the same beginning, like MO_1 and MO_10, the qc.and.removeDoublets
function isn't able to show the table with deleted events...
barplot.cell.pops
-> default option for color.by
(avoid errors).tree
methods to our code for avoiding dependences.surv.tree
not correctly exported.maxstat
.dim.reduction
(and change tumap
by umap
?)fsom.metaclustering
Specify different name for name it if not X11
In data.table::melt(table(mnames)) :
The melt generic in data.table has been passed a table and will attempt to redirect to the relevant reshape2 method; please note that reshape2 is deprecated, and this redirection is now deprecated as well. To continue using melt methods from reshape2 while both libraries are attached, e.g. melt.list, you can prepend the namespace like reshape2::melt(table(mnames)). In the next version, this warning will become an error.
Collapse channel::marker to see if discrepancies or identical marker names (Irene's request).
Calculate the median and highlight those patients with a SD higher than X...
Divergent colors function... and link to colors_palette.
barplot.cell.pops_m <- function (cell_clusters, metadata, colname_sampleID, return_table)
{
tab <- table(cell_clusters, metadata[,colname_sampleID])
prop_tableL <- prop.table(tab, margin = 2) * 100
ggdfL <- data.table::melt(prop_tableL, value.name = "proportion")
colnames(ggdfL)[2] <- colname_sampleID
mmL <- match(ggdfL[, colname_sampleID], metadata[, colname_sampleID])
ggdfL <- data.frame(metadata[mmL, ], ggdfL)
g <- ggplot(ggdfL, aes_string(x = "sample_id", y = "proportion",
fill = "cell_clusters")) + geom_bar(stat = "identity") +
facet_wrap(~condition, scales = "free_x") + theme_minimal() +
theme(axis.text.x = element_text(angle = 90, hjust = 1)) +
scale_fill_manual(values = colors_palette)
if ("try-error" %in% class(suppressWarnings(try(x11(), silent = T)))) {
cat("X11 is not active, boxplot is saved in -> barPlot_cell_prop.",
deparse(substitute(cell_clusters)), ".jpg\n", sep = "")
suppressMessages(ggsave(paste0("barPlot_cell_prop.",
deparse(substitute(cell_clusters)), ".jpg"), device = "jpeg",
plot = g))
}
else {
print(g)
}
if(return_table == "counts"){
return(tab)
}else{ #add "percentage" option and block any other possibilities
return(prop_tableL)
}
}
Si se ha hecho la reducción o se han seleccionado una serie de eventos previamente (ej, LiB) hay que indicarle que no hay reduction_suffix y omitir la eliminación de los dobletes (si se han eliminado manualmente, luego podrían perderse eventos interesantes).
>qc.and.removeDoublets(reduction_suffix="")
Processing: Screening_013543_TubeT.fcs
Processing: Screening_014090_TubeT.fcs
Processing: Screening_014184_TubeT.fcs
Processing: Screening_014408_TubeT.fcs
Processing: Screening_014497_TubeT.fcs
Processing: Screening_014786_TubeT.fcs
Processing: Screening_015308_TubeT.fcs
Processing: Screening_015879_TubeT.fcs
Processing: Screening_016286_TubeT.fcs
Processing: Screening_016469_TubeT.fcs
Processing: Screening_016692_TubeT.fcs
Processing: Screening_017425_TubeT.fcs
Processing: Screening_017499_TubeT.fcs
Processing: Screening_017529_TubeT.fcs
Processing: Screening_017557_TubeT.fcs
Processing: Screening_017759_TubeT.fcs
Processing: Screening_017789_TubeT.fcs
Processing: Screening_018108_TubeT.fcs
Processing: Screening_018769_TubeT.fcs
Processing: Screening_018796_TubeT.fcs
Processing: Screening_019278_TubeT.fcs
Processing: Screening_019649_TubeT.fcs
Processing: Screening_020616_TubeT.fcs
Processing: Screening_020653_TubeT.fcs
Processing: Screening_021098_TubeT.fcs
Error in cat("WARNING! >", i, "has lost some much cells (more that 30%) in the QC and doublets removal steps, consider to review it!", :
argument is missing, with no default
In addition: Warning message:
In dir.create(output_folder) : 'results_preprocessing' already exists
I just modified the div.colors function by taking advantage of the colorspace package. Please look at the function and think on the possibility of replacing the old one with this one (Why the 2nd part depends upon a number >74?). Attached you'll find an UMAP with this automatically generated qualitative palette.
div.colors <- function(n, set.seed = 333){
require(RColorBrewer)
require(colorspace)
if(n < 74){
col<-qualitative_hcl(n, palette = "Dark3")
}else{
qual_col_pals = sample(brewer.pal.info)
col_vector = sample(unlist(mapply(brewer.pal, qual_col_pals$maxcolors, rownames(qual_col_pals))))
col = sample(col_vector, 333)
set.seed(set.seed); col <- sample(col, n, replace = T)
}
return(col)
}
Here you'll find my code with result obtained!
df<-as.data.frame(t(assay(fcs1000,"logcounts")))
df_expr<-cbind(df, SOM=fcs1000$SOM)
#or if only selected clusters
df_expr<-df_expr[df_expr$SOM==c("1","2","3","4"),]
# prepare the dotplot
pmain<-ggplot(df_expr, aes(x=CD4, y=CD8, color=SOM)) +
geom_point(size=0.5)+
scale_color_manual(values = color_clusters) +
theme_bw()
###if density plot needed
#ggplot(df_expr, aes(x=CD4, y=CD8)) +
# geom_point(size=0.5)+theme_bw() +geom_hex(bins = 200) +
# scale_fill_continuous(type = "viridis")
library(cowplot)
# histogram along x axis
xdens <- axis_canvas(pmain, axis = "x")+
scale_fill_manual(values = color_clusters) +
geom_density(data = df_expr, aes(x = CD4, fill = SOM),
alpha = 0.7, size = 0.2)
# histogram along y axis
ydens <- axis_canvas(pmain, axis = "y", coord_flip = TRUE)+
scale_fill_manual(values = color_clusters) +
geom_density(data = df_expr, aes(x = CD8, fill = SOM),
alpha = 0.7, size = 0.2)+
coord_flip()
p1 <- insert_xaxis_grob(pmain, xdens, grid::unit(.2, "null"), position = "top")
p2<- insert_yaxis_grob(p1, ydens, grid::unit(.2, "null"), position = "right")
ggdraw(p2)
I added a new assay.i named "scaled" with scaled data....it should be used along the code as default to be sure to use the correct set of data..
Maybe some default parameters should be modified!
scale.data <- function(fcs.SE, assay.i = "normalized",scaled.matrix.name = "scaled"){
data <- t(assay(fcs.SE, i = assay.i))
rng <- matrixStats::colQuantiles(data, probs = c(0.01, 0.99))
expr01 <- t((t(data) - rng[, 1]) / (rng[, 2] - rng[, 1]))
expr01[expr01 < 0] <- 0
expr01[expr01 > 1] <- 1
SummarizedExperiment::assay(fcs.SE, i = scaled.matrix.name) <-t(expr01)
return(fcs.SE)
}
pb <- txtProgressBar(min = 0, max = nrow(gene.sum),
style = 3)
for (i in 1:nrow(gene.sum)) {
prot.dat = all.prot.dat[Hugo_Symbol %in% gene.sum[i,
Hugo_Symbol]]
syn.res = rbind(syn.res, cluster_prot(prot.dat = prot.dat,
gene = gene.sum[i, Hugo_Symbol], th = gene.sum[i,
th], protLen = gene.sum[i, aa.length]))
setTxtProgressBar(pb, i)
}
(based in maftools::oncodrive::parse_prot)
densityplot(as.formula(paste0("~", marker)), datr, main = "normalized", xlim = lims.FCS(fcs), filter=curv1Filter(marker), legend = F)
Error: $ operator is invalid for atomic vectors
density does not allow to plot individual markers???
When I try to set array.i ="transformed" I have this error:
Error in assay(fcs.SCE, i = assay.i) :
'assay(, i="character", ...)' invalid subscript 'i'
'normalized' not in names(assays())
Maybe we can applied this method in FlowCT?? Take a look!
This is the error...no problem if I download and install from tar.gz
we should look at all dependencies...
Error: (converted from warning) package 'Hmisc' was built under R version 4.0.2
Execution halted
ERROR: lazy loading failed for package 'FlowCT.devel'
EDIT: this error is not reproducible...I changed computer and it does not appear...For sure not related to FlowCT!
It would be useful to introduce the ggfortify package (https://cran.r-project.org/web/packages/ggfortify/vignettes/plot_pca.html) to "cluster" all the different patients/samples (even at the subclustering step!)...in this way we can introduce a new variable that could be useful for downstream analysis!
A declarative, efficient, and flexible JavaScript library for building user interfaces.
🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
An Open Source Machine Learning Framework for Everyone
The Web framework for perfectionists with deadlines.
A PHP framework for web artisans
Bring data to life with SVG, Canvas and HTML. 📊📈🎉
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
Some thing interesting about web. New door for the world.
A server is a program made to process requests and deliver data to clients.
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
Some thing interesting about visualization, use data art
Some thing interesting about game, make everyone happy.
We are working to build community through open source technology. NB: members must have two-factor auth.
Open source projects and samples from Microsoft.
Google ❤️ Open Source for everyone.
Alibaba Open Source for everyone
Data-Driven Documents codes.
China tencent open source team.