Dimensionality reduction and clustering

Approximated t-SNE

Description

https://github.com/tercen/atsne_docker_operator.git

Build

VERSION=1.1.11
docker build -t tercen/atsne:$VERSION .
docker push tercen/atsne:$VERSION
# see operator.json :-- "container": "tercen/atsne:1.1.8"
git add -A && git commit -m "$VERSION" && git tag -a $VERSION -m "++" && git push && git push :--tags

# renv cache ~/.local/share/renv
docker run -it :--rm :--entrypoint "/bin/bash" tercen/atsne:1.1.11 

GitHub link

atsne_docker_operator on GitHub

Clustering metrics operator

Description

clustering_metrics operator returns clustering metrics.

Usage

Input projection	.
row	represents the variables
col	represents the observations
label	represents the clusters
y-axis	is the value of the measurement

Output relations	.
metrics	character, name of the clustering metric
value	numeric, value of the clustering metric

Details

References

This operator is based on the clusterCrit R function.

See Also

hierarchical_clustering_operator, flowsom_operator, clusterx_operator

GitHub link

clustering_metrics_operator on GitHub

Hierarchical clustering tree operator

Description

clustering_tree operator returns a hierarchical clustering tree to be projected in Tercen.

Usage

Input projection	.
row	factor, variables to cluster
col	factor, variables to cluster (dist_to variable from a dist operator)
y-axis	numeric, pairwise distance (dist variable from a dist operator)

Output relations	.
presence	numeric, to be projected on y-axis
tree_dim1	factor, to be projected on rows
tree_dim2	factor, to be projected on columns
tip_labels	factor, leaf labels, to be projected on rows

GitHub link

clustering_tree_operator on GitHub

clusterx operator

Description

clusterx operator performs a fast clustering by automatic search and find of density peaks

Usage

Input projection	.
row	represents the variables (e.g. channels, markers)
col	represents the observations (e.g. cells, samples, individuals)
x-axis	first axis
y-axis	second axis

Input parameters	.
dimReduction	type of reduction to perform, pca, tsne, NULL, default is NULL
outDim	number of demensions to return, default 2

Output relations	.
cluster	character, returns a cluster id per value, per cell

Details

clusterx operator performs a fast clustering by automatic search and find of density peaks.

References

see https://github.com/JinmiaoChenLab/ClusterX

See Also

clusterx

Examples

GitHub link

clusterx2_operator on GitHub

clusterx operator

Description

clusterx operator performs a fast clustering by automatic search and find of density peaks

Usage

Input projection	.
row	represents the variables (e.g. channels, markers)
col	represents the observations (e.g. cells, samples, individuals)
y-axis	is the measurement value

Input parameters	.
dimReduction	type of reduction to perform, pca, tsne, NULL, default is NULL
outDim	number of demensions to return, default 2

Output relations	.
cluster	character, returns a cluster id per column (e.g. per cell)

Details

clusterx operator performs a fast clustering by automatic search and find of density peaks.

References

see https://github.com/JinmiaoChenLab/ClusterX

See Also

clusterx

Examples

GitHub link

clusterx_operator on GitHub

Fast t-SNE Docker Operator

Build the image

VERSION=0.0.1
docker build -t tercen/fast_tSNE_docker_operator:$VERSION .
docker push tercen/fast_tSNE_docker_operator:$VERSION
git add -A && git commit -m "$VERSION" && git tag  $VERSION  && git push && git push :--tags

GitHub link

fast_tSNE_docker_operator on GitHub

Fast t-SNE operator

Description

The Fast t-SNE operator performs the Fast Fourier Transform Interpolation-based t-SNE dimensionality reduction method.

Usage

Input projection	.
row	represents the variables (e.g. genes, channels, markers)
col	represents the observations (e.g. cells, samples, individuals)
y-axis	measurement value

Output relations	.
tsne1, tsne2	first two components containing the new projected values

Details

The operator performs tSNE analysis. It reduces the amount of variables (i.e. indicated by rows) to a lower number (default 2).

Reference

FFT-accelerated Interpolation-based t-SNE

See Also

tsne pca

GitHub link

fast_tSNE_operator on GitHub

hclust operator

Description

hclust operator performs a hierarchical clustering.

Usage

Input projection	.
row	represents the row data
col	represents the col data
y-axis	is the value of measurement

Input parameters	.
scale	boolean, scaled to have unit variance before the analysis takes place
center	boolean, shifted to be zero center before the analysis takes place
fill	numeric, a fill in value for datapoints structural missings

Output relations	.
rorder	numeric, order of rows after clustering
corder	numeric, order of cols after clustering

Details

The operator is the hclust function of the base R package.

References

see https://en.wikipedia.org/wiki/Hierarchical_clustering

See Also

Examples

GitHub link

hclust_operator on GitHub

MDS operator

Description

MDS operator performs a Multidimensional Scaling analysis.

Usage

Input projection	.
y-axis	numeric, distance measure
col	character, dist_to variable obtained from a pairwise_distance operator
row	character, variables

Output relations	.
mds_1	numeric, first dimension
mds_2	numeric, second dimension

Details

The operator takes as input a pariwise distance matrix as obtained with the pairwise_distance_operator.

References

This operator is a wrapper of the cmdsale R function.

See Also

pca_operator, pairwise_distance_operator

GitHub link

mds_operator on GitHub

flowsom operator

Description

flowsom operator performs the SOM (self organizing maps) in the flowSOM R package.

Usage

Input projection	.
row	represents the variables (e.g. channels, markers)
col	represents the observations (e.g. cells)
y-axis	is the value of measurement signal of the channel/marker

Input parameters	.
xdim	Width of the grid
ydim	Hight of the grid
rlen	Number of times to loop over the training data for each MST
mst	Number of times to build an MST
alpha_start	Start learning rate
alpha_end	End learning rate
dstf	Distance function (1=manhattan, 2=euclidean, 3=chebyshev, 4=cosine)

Output relations	.
mapping_node_num	numeric, per column (e.g. per cell)
mapping_node_label	character, per column (e.g. per cell)

Details

The operator is the SOM function of the flowSOM R package.

References

see the flowSOM::SOM function of the R package for the documentation,

See Also

clusterx

Examples

GitHub link

oldflowsom_operator on GitHub

pca operator

Description

pca operator performs principle component analysis.

Usage

Input projection	.
row	represents the variables (e.g. genes, channels, markers)
col	represents the observations (e.g. cells, samples, individuals)
y-axis	measurement value

Input parameters	.
scale	logical, indicating whether the variables should be scaled to have unit variance before the analysis takes place
center	logical, indicating whether the variables should be shifted to be zero centered before the analysis takes place
na.action	A function which indicates what should happen when the data contain NAs
tol	numeric, indicating the magnitude below which components should be omitted. Components are omitted if their standard deviations are less than or equal to tol times the standard deviation of the first component
maxComp	numeric, maximum number of components to return, default 5

Output relations	.
pca1, pca2, pca3, pca4, pca5	first five components containing the new projected values

Details

The operator performs principal component analysis. It reduces the amount of variables (i.e. indicated by rows) to a lower number (default 5).

Reference

See Also

tsne

Examples

GitHub link

pca_operator on GitHub

rphenograph operator

Description

rephenograph operator performs a phenotype clustering in the Rphenograph R package.

Usage

Input projection	.
row	represents the variables (e.g. channels, markers)
col	represents the observations (e.g. cells)
y-axis	is the value of measurement signal of the channel/marker

Input parameters	.
xdim	Width of the grid
ydim	Hight of the grid
rlen	Number of times to loop over the training data for each MST
mst	Number of times to build an MST
alpha_start	Start learning rate
alpha_end	End learning rate
dstf	Distance function (1=manhattan, 2=euclidean, 3=chebyshev, 4=cosine)

Output relations	.
mapping_node_num	numeric, per column (e.g. per cell)
mapping_node_label	character, per column (e.g. per cell)

Details

The operator is the rphenograph function of the Rphenograh R package.

References

see the rphenograph::SOM function of the R package for the documentation,

See Also

Examples

GitHub link

rphenograph_operator on GitHub

somflow operator

Description

somflow operator performs the SOM (self organizing maps) in the FlowSOM R package.

Usage

Input projection	.
row	represents the variables (e.g. channels, markers)
col	represents the observations (e.g. cells)
y-axis	is the value of measurement signal of the channel/marker

Input parameters	.
xdim	Width of the grid
ydim	Hight of the grid
rlen	Number of times to loop over the training data for each MST
mst	Number of times to build an MST
alpha_start	Start learning rate
alpha_end	End learning rate
dstf	Distance function (1=manhattan, 2=euclidean, 3=chebyshev, 4=cosine)

Output relations	.
mapping_node_label	character, per column (e.g. per cell)

Details

The operator is the SOM function of the flowSOM R package.

References

see the FlowSOM::SOM function of the R package for the documentation,

See Also

clusterx

Examples

GitHub link

somflow_operator on GitHub

tsne operator

Description

tsne operator performs tSNE analysis.

Usage

Input projection	.
row	represents the variables (e.g. genes, channels, markers)
col	represents the observations (e.g. cells, samples, individuals)
y-axis	measurement value

Input parameters	.
dims	logical, output dimensionality, default 2
initial_dims	numeric, the number of dimensions that should be retained in the initial PCA step, default 50
perplexity	numeric, perplexity parameter, default is 30
theta	numeric, speed/accuracy trade-off (increase for less accuracy), set to 0.0 for exact TSNE, default 0.05
pca	numeric, whether an initial PCA step should be performed, default TRUE
max_iter	numeric, number of iteration, default 1000
pca_center	logical, should data be centered before pca is applied ?
pca_scale	logical, should data be scaled before pca is applied ?
stop_lying_iter	numeric, Iteration after which the perplexities are no longer exaggerated
mom_switch_iter	numeric, Iteration after which the final momentum is used

Output relations	.
tsne1, tsne2	first two components containing the new projected values

Details

The operator performs tSNE analysis. It reduces the amount of variables (i.e. indicated by rows) to a lower number (default 2).

Reference

See Also

pca

Examples

GitHub link

tsne_operator on GitHub