error message

[kasumaz]

HI, I am using PhateR. It looks very neat. I can get the vignette going and all working smooth. When I try with my data and run:

scex_PCA <- as.data.frame(prcomp(scex)$x) Error in colMeans(x, na.rm = TRUE) : 'x' must be numeric

With the above error message, i went back to the matrix and I cant see something wrong with it. Would it be worth while to have a look at your end? I start with a smaller bunch of cells of about 130 cells. Thanks a lot. scext.csv.zip

[scottgigante]

Hi @kasumaz , it looks like the file you are reading is not comprised entirely of numbers. I was able to run prcomp on the data you provided by loading it as follows:

library(tidyverse)
df <- read_csv("scext.csv", col_types =list(V1=col_character(), .default=col_double())) %>% 
  column_to_rownames("V1")
prcomp(df)

[kasumaz]

Hi, Thanks for your reply. We you able to proceed with any other of the commands for PhateR?
Just wondering what else I should do aside from checking the matrix thoroughly. I was using the same Matrix in seurat. But I will go over it now.

[scottgigante]

I just ran phateR::phate(df) after the above commands and it worked fine. Let me know if you have further problems. Passing .default=col_double() to read_csv should raise warnings if there are any other non-numeric entries.

[kasumaz]

Would you be able to send me over the script you used if you still have it and I try at my end? Maybe you have something slightly different from the tutorial that I give it a go. I will check out otherwise. Apologies, I am new to R but getting there. Thanks.

[scottgigante]

No problem.

library(phateR)
library(tidyverse)
library(viridis)
library(Rmagic)

# read data
df <- read_csv("scext.csv", col_types =list(V1=col_character(), .default=col_double())) %>% 
  column_to_rownames("V1")

# keep genes expressed in at least 10 cells
keep_cols <- colSums(df > 0) > 10
df <- df[,keep_cols]

# keep cells with at least 1000 UMIs
keep_rows <- rowSums(df) > 1000
df <- df[keep_rows,]

# normalize and transform
df <- library.size.normalize(df)
df <- sqrt(df)

# PCA
df_pca <- as.data.frame(prcomp(df)$x)
ggplot(df_pca) +
  geom_point(aes(PC1, PC2, color=df$Chd4)) +
  labs(color="Chd4") +
  scale_color_viridis(option="B")

df_phate <- phate(df)

ggplot(df_phate) +
  geom_point(aes(PHATE1, PHATE2, color=df$Chd4)) +
  labs(color="Chd4") +
  scale_color_viridis(option="B")

This worked on the sample you sent me.

[kasumaz]

Hi mate. Thanks soo much. This now works at my end. I will use it now. If you have any other tools in the pipeline/developing that are similar or an extension, let me know and I can implement them in. I will keep a look out otherwise. Thanks again.

[scottgigante]

Glad to hear it's working, and thanks very much for the offer to contribute! We always welcome PRs to our codebase if you identify somewhere you can help us improve :)