Component: MMClustering

MMClustering

Normal, t, Skew-normal, and skew-t Mixture-Model Clustering.

Version	1.0
Bundle	flowand
Categories	FlowCytometry
Authors	Ali Oghabian (ali.oghabian@helsinki.fi)
Requires	R, Rmpi
Source files	component.xml MMClustering.r
Usage	Example with default values MMClustering( fcsFiles = CSVList, IDCol = "", NaRmove = true, channelsToCluster = <string>, channelsToRegularPlot = "", clusterIDColName = "cluster", density = "skewt", epsilon = 0.0001, estimateMode = false, gMax = 5, gMin = 3, includePosterior = true, iterationMaximum = 5000, nSlaves = 5, pagebreak = false, pch = ".", plotType = "boundry2d", randomSeed = 1984, step = 0.5, writeImage = false, writeParams = false )

Version

1.0

Bundle

flowand

Categories

FlowCytometry

Authors

Ali Oghabian (ali.oghabian@helsinki.fi)

Requires

R, Rmpi

Source files

component.xml MMClustering.r

Usage

Example with default values

Name	Type	Mandatory	Description
fcsFiles	CSVList	Mandatory	Location of the csv files to be clustered e.g. Preprocessed files in flow cytometry analysis .

Name

Type

Mandatory

Description

fcsFiles

CSVList

Mandatory

Location of the csv files to be clustered e.g. Preprocessed files in flow cytometry analysis .

Outputs

Name	Type	Description
clusters	CSVList	csv files including the input data and one additional column presenting the mixture modeling cluster results. The numbers added to the input filenames represent the g parameter number.
postProb	CSVList	Posterior probabilities of the data. For each file in fcsFiles generates a file with similar name ehich includes the psterior probability of each point of the data. The dimentions of each csv file is similar to dimentions of the clsuetred data i.e. (size of the ChannelsToCluster)*( Row number of the data).
report	Latex	Visualization of MixtureModeling results.
mmSpecs	CSV	The mmSpecs.csv file, include the information and calculations for the Mixture Modeling Clustering that are suitable for OptimalClustering component. The first column includes the file names and the other columns are the aic, bic and SWR calculations, which could be used to estimate the best clustering g parameter
mmParams	BinaryFolder	.txt files including parameters for the Mixture model results. They are in R object format, that have been presented in an R scripts text.

Name

Type

Description

clusters

CSVList

csv files including the input data and one additional column presenting the mixture modeling cluster results. The numbers added to the input filenames represent the g parameter number.

postProb

CSVList

Posterior probabilities of the data. For each file in fcsFiles generates a file with similar name ehich includes the psterior probability of each point of the data. The dimentions of each csv file is similar to dimentions of the clsuetred data i.e. (size of the ChannelsToCluster)*( Row number of the data).

report

Latex

Visualization of MixtureModeling results.

mmSpecs

CSV

The mmSpecs.csv file, include the information and calculations for the Mixture Modeling Clustering that are suitable for OptimalClustering component. The first column includes the file names and the other columns are the aic, bic and SWR calculations, which could be used to estimate the best clustering g parameter

mmParams

BinaryFolder

.txt files including parameters for the Mixture model results. They are in R object format, that have been presented in an R scripts text.

Parameters

Name	Type	Default	Description
IDCol	string	""	The column representing the row numbers. The parameter is optional and it would be added to the "postProb" files, if provided.
NaRmove	boolean	true	Whether remove the rows with na from the data or return an Error.
channelsToCluster	string	(no default)	A comma-separated list of column names or column numbers (e.g. channel names/ numbers in flow cytometry analysis), to be clustered. As for instance the value could be set as "1, 2, 3, 4, 5" or "FSC.A, SSC, ERK1, STAT1, CD4" .
channelsToRegularPlot	string	""	A comma-separated list of column names or column numbers (e.g. channel names/ numbers in flow cytometry analysis), to be plotted. As for instance the value could be set as "1, 2, 3, 4, 5" or "FSC.A, SSC, ERK1, STAT1, CD4". This parameter is only for the "regular2d" and "regullar3d" plottings.
clusterIDColName	string	"cluster"	The name of the column in the `clusters` output which contains the cluster IDs of the rows.
density	string	"skewt"	Density distribution to be used for clustering: normal, t, skew normal, skew t.
epsilon	float	0.0001	A value used for threshold in the mixture modeling fitting process. The smaller the value the more accurate the cluster results are. If the clusters do not satisfy the epsilon threshold in iterationMaximum iteration attempts an error would be returned.
estimateMode	boolean	false	Used only for skew distributions. Whether to estimate the mode for each cluster. The defalt value (i.e. false) is recommended
gMax	int	5	The maximum number of the "Mixture Model" clusters. It is considering that results for some clusters might be Null. As for instance, the results for MMclustering with g parameter (i.e. expected cluster number) of 5 might group the rows of the data in 4 clusters of 1,2,4,and 5.
gMin	int	3	The minimum number of the Mixture Model clusters.It is considering that results for some clusters might be Null. As for instance, the results for MMclustering with g parameter (i.e. expected cluster number) of 5 might group the rows of the data in 4 clusters of 1,2,4,and 5.
includePosterior	boolean	true	Whether include the Posterior probabilities. Check "the postProb" output for more information.
iterationMaximum	int	5000	The maximum limitation for the number of iterations, in the mixturemodeling fitting process
nSlaves	int	5	The number of the clusters used for paralelising the Mixture Modeling.
pagebreak	boolean	false	Tells if the result document should start with a page break.
pch	string	"."	The type and shape of plotting points. Numbers (e.g. "2") and characters (e.g. ".") are possible.
plotType	string	"boundry2d"	The type of Plotting. The choices are regular2d, regular3d, contour2d, individualcontour2d, and boundry2d.
randomSeed	int	1984	The seed value for semi-random processes
step	float	0.5	Only used when Density is "skew" and the estimateMode is "true". It is used to calculated the number of the iterations of the Expected-Maximization (i.e. EM), while computing the Maximum Likelihood (i.e ML) estimate. The value should be greater than 0. The smaller the value, the more accurate the estimations would be.
writeImage	boolean	false	Whether include the plotting images in the report directory.
writeParams	boolean	false	Whether write the Mixture Model parameters.

Name

Type

Default

Description

IDCol

string

The column representing the row numbers. The parameter is optional and it would be added to the "postProb" files, if provided.

NaRmove

boolean

true

Whether remove the rows with na from the data or return an Error.

channelsToCluster

string

(no default)

A comma-separated list of column names or column numbers (e.g. channel names/ numbers in flow cytometry analysis), to be clustered. As for instance the value could be set as "1, 2, 3, 4, 5" or "FSC.A, SSC, ERK1, STAT1, CD4" .

channelsToRegularPlot

string

A comma-separated list of column names or column numbers (e.g. channel names/ numbers in flow cytometry analysis), to be plotted. As for instance the value could be set as "1, 2, 3, 4, 5" or "FSC.A, SSC, ERK1, STAT1, CD4". This parameter is only for the "regular2d" and "regullar3d" plottings.

clusterIDColName

string

"cluster"

The name of the column in the clusters output which contains the cluster IDs of the rows.

density

string

"skewt"

Density distribution to be used for clustering: normal, t, skew normal, skew t.

epsilon

float

0.0001

A value used for threshold in the mixture modeling fitting process. The smaller the value the more accurate the cluster results are. If the clusters do not satisfy the epsilon threshold in iterationMaximum iteration attempts an error would be returned.

estimateMode

boolean

false

Used only for skew distributions. Whether to estimate the mode for each cluster. The defalt value (i.e. false) is recommended

gMax

int

The maximum number of the "Mixture Model" clusters. It is considering that results for some clusters might be Null. As for instance, the results for MMclustering with g parameter (i.e. expected cluster number) of 5 might group the rows of the data in 4 clusters of 1,2,4,and 5.

gMin

int

The minimum number of the Mixture Model clusters.It is considering that results for some clusters might be Null. As for instance, the results for MMclustering with g parameter (i.e. expected cluster number) of 5 might group the rows of the data in 4 clusters of 1,2,4,and 5.

includePosterior

boolean

true

Whether include the Posterior probabilities. Check "the postProb" output for more information.

iterationMaximum

int

5000

The maximum limitation for the number of iterations, in the mixturemodeling fitting process

nSlaves

int

The number of the clusters used for paralelising the Mixture Modeling.

pagebreak

boolean

false

Tells if the result document should start with a page break.

pch

string

"."

The type and shape of plotting points. Numbers (e.g. "2") and characters (e.g. ".") are possible.

plotType

string

"boundry2d"

The type of Plotting. The choices are regular2d, regular3d, contour2d, individualcontour2d, and boundry2d.

randomSeed

int

1984

The seed value for semi-random processes

step

float

0.5

Only used when Density is "skew" and the estimateMode is "true". It is used to calculated the number of the iterations of the Expected-Maximization (i.e. EM), while computing the Maximum Likelihood (i.e ML) estimate. The value should be greater than 0. The smaller the value, the more accurate the estimations would be.

writeImage

boolean

false

Whether include the plotting images in the report directory.

writeParams

boolean

false

Whether write the Mixture Model parameters.

Anduril

MMClustering

Inputs

Outputs

Parameters