Wiki

Exporting data

Saving the charts or tables

Data from all charts and tables can be copied or saved by right clicking, and selecting either Copy data or Export data. In either case, the underlying data will be copied in tab separated format, suitable for pasting into Excel or other software.

The charts can be saved as images via right clicking and Save as.... When it comes to making pretty figures for papers though, I usually recommend exporting the measurements and drawing your own.

Exporting all measurements

To export raw data for single or multiple datasets, select the datasets you want to export, and choose Dataset > Export...

There are several options available depending on what data you need. If a single dataset is selected, the measurements will be exported to a file of your choice. By default, the file will be in the dataset's folder: e.g. if the dataset is saved to C:\folder\2017-04-11_12-00-00\Dataset.nmd then the default output file will be C:\folder\2017-04-11_12-00-00\Dataset.txt

If multiple datasets are selected, the combined measurements will be exported to a file of your choice. By default, the file will be called Multiple_stats_export.txt.

Columns are usually separated by tabs.

Option	Action
Nuclear measurements	The measured values for each nucleus, plus normalised profiles
Full nuclear profiles	The non-normalised angle profile values for each nucleus. Since there are a different number of values in each nucleus, they are combined into a single column, separated by commas for later parsing
Full nuclear outlines	The X and Y coordinates of each border point in the outline of each nucleus. Since there are a different number of values in each nucleus, they are combined into a single column. XY pairs are separated by commas for later parsing. Within each pair, X and Y values are separated by a pipe (`\|`) e.g. `12.4\|3.6,12.9\|4.0`.
Nuclear signal measurements	The measured values for each nuclear signal
Nuclear signal shells	The measured nuclear signal in each shell, following shell analysis
Cell locations within images	The XY coordinates of nucleus centres-of-mass within their source image
Consensus nuclei as SVG	The consensus nucleus outlines for all selected datasets in SVG so you can make custom figures
Single cell images	Export each nucleus as a separate image to a folder in the same directory as the `nmd` file. You will be given the option to mask out background; if this is selected, any pixels not within the nucleus will be set to black.
Dataset analysis options	The options used to analyse this dataset in XML format. This can be used directly to set up subsequent analyses
Landmark rulesets	The rules used to detect landmarks, so you can modify them for custom nucleus types

Running downstream analysis on exported data

Once you have exported the data, you can do whatever downstream analysis you wish. This demonstrates a simple example of how to run a tSNE in R, and perform a hierarchical clustering on the result, using the angle profiles from exported nuclear statistics.

# Demonstration of how to read a sample tsv of morphology data, extract the relevant
# columns for the angle profile, run a tSNE, cluster, and display the result
library(tidyverse)
library(Rtsne)
library(cluster)
library(dendextend)

# Read in the input data file
data <- read.csv("Exported_data.tsv", sep="\t", header=T, stringsAsFactors = F)

# Take just the angle profile columns
profiles <- data %>% dplyr::select(one_of(paste0("Angle_profile_", seq(0,99,1))))

# Set the rownames of the data using the cell id and the strain for convenience
rownames(profiles) <- data$CellID

# Set the random number generator with a seed for reproducilble results
set.seed(42)

# Run a tSNE on the profile data - the Rtsne function requires a matrix
# Perplexity should be tuned according to the number of samples you have;
# a good starting point is 5% of your sample size
rtsne_out <- Rtsne(as.matrix(profiles), perplexity=20, max_iter=1000)
tsne.values <- as.data.frame(rtsne_out$Y)

# Display the results as a scatter plot
ggplot(tsne.values, aes(x=V1, y=V2))+
  geom_point()

# Cluster the tSNE values using hierarchical clustering
# The agnes function is agglomerative nesting
hc <- cluster::agnes(tsne.values, method = "ward")

# Make a dendrogram (tree) from the cluster data
dend <- as.dendrogram(hc)

# Cut the dendrogram to get 4 separate groups
clusters <- dendextend::cutree(dend, 4)

# Assign the group names to the original data
data$agnes <- clusters

# Assign the group names to the tSNE results
tsne.values$agnes <- clusters

# Plot the tSNE results coloured by cluster
ggplot(tsne.values, aes(x=V1, y=V2, col=agnes))+
geom_point()

This would also be suitable for generating a cluster map to import back into the program