| Title: | An R package for flow data clustering |
|---|---|
| Description: | A fast and automatic clustering to classify the cells into subpopulations based on finding the peaks from the overall density function generated by K-means. |
| Authors: | Yongchao Ge<[email protected]> |
| Maintainer: | Yongchao Ge<[email protected]> |
| License: | Artistic-1.0 |
| Version: | 1.53.2 |
| Built: | 2024-11-22 03:06:39 UTC |
| Source: | https://github.com/bioc/flowPeaks |
Adjusting the smoothing and merging behavior of the flowPeaks results by changing the multiplers of the covariance matrix and the tolerance level for joining two peaks
adjust.flowPeaks(object,tol,h0,h,...)adjust.flowPeaks(object,tol,h0,h,...)
object |
The output from the function |
tol |
See |
h0 |
See |
h |
See |
... |
Optional additional arguments. At present no additional arguments are used. |
It returns an updated object of class flowPeaks, the detail defintion
of which
can be seen in
flowPeaks.
Yongchao Ge [email protected]
The function takes a flowPeaks output and a new data set (or could be the same dataset that generated the flowPeaks), and compute the cluster label assignment
assign.flowPeaks(fp,A,tol=0.01,fc=0.8)assign.flowPeaks(fp,A,tol=0.01,fc=0.8)
fp |
an object of class flowPeaks, the output from the function
|
A |
A data matrix with the same number of columns as the data that geneterated fp |
tol |
All points where the probability density is less than tol (default is 1%) of the peak denisty of that cluster are labled as outliers. If tol is set 0, no outliers according to this rule. The details can be seen in the first equation of Section 2.5 in the flowPeaks manuscript (Ge et al 2012) |
fc |
All points where the classified cluster contributes less than fc (default is 80%) of overall denstiy are labeled as outliers. if fc is set to 0%, no outliers can be found according to this rule. The details can be seen in the second equation of Section 2.5 in the flowPeaks manuscript (Ge et al 2012) |
It returns the class label assignment of each data point, where -1 indicates outliers. When A is the same data that generated fp, If tol is 1 and fc is 0, the returned labels are the same as fp$peaks.cluster.
Yongchao Ge [email protected]
Ge Y. et al, flowPeaks: a fast unsupervised clustering for flow cytometry data via K-means and density peak finding, 2012, Bioinformatics, 8(15):2052-8
A flow cytometry data that is used barcode to measure many samples simultaneusly
data(barcode)data(barcode)
An object (barcode) of data frame with 180912 rows and 3 columns and a vector (barcode.cid) for the cluster labels accoring to the manual gating.
The data is a random subset of the full data set for Figure 3A of the paper (Sugar et al 2010), This subset was used to do all comparisons in the paper (Ge et al 2012) with other clustering algorithms.
Sugar I. P. and Sealfon S. C., Misty Mountain clustering:
application to fast unsupervised flow cytometry gating,
BMC Bioinformatcs, 2010, 11:502.
Ge Y. et al, flowPeaks: a fast unsupervised clustering for flow cytometry data via K-means and density peak finding, 2012, Bioinformatics 8(15):2052-8
A simulated flowcytometry data with two concave shapes
data(concave)data(concave)
An object (concave) of data frame with rows and 3 columns and a vector (concave.cid) for the true cluster labels.
Ge Y. et al, flowPeaks: a fast unsupervised clustering for flow cytometry data via K-means and density peak finding, Bioinformatics 8(15):2052-8
This function takes the cluster labels of the two clusterings, one is based on the gold standard, the other is a candidate clusterign, and compute one of the three metrics to assess the candidate clustering performance.
evalCluster(gs,cand,method=c("Rand.index","Fmeasure","Vmeasure"), rm.gs.outliers=TRUE)evalCluster(gs,cand,method=c("Rand.index","Fmeasure","Vmeasure"), rm.gs.outliers=TRUE)
gs |
A integer-valued vector of length n for the cluster labels of the gold standard clustering, where negative numbers such as -1 is for the outerliers |
cand |
A integer-valued vector of length n for the cluster label of a candidate clustering, where -1 is for the outliers |
rm.gs.outliers |
Determining whether the outliers of the gold standard clustering should be removed in the comparison |
method |
A single character to indicate which one of three metrics
should be used to evaluate the clustering. The details are described in
Ge (2012) and references mentioned in that paper
|
Yongchao Ge [email protected]
Ge Y. et al, flowPeaks: a fast unsupervised clustering for flow cytometry data via K-means and density peak finding, 2012, Bioinformatics 8(15):2052-8
This is the core function in the flowPeaks package. It
generates the output of the cluster and information associated with
each cluster, which can be used by the function plot for
visualization
flowPeaks(x,tol=0.1,h0=1,h=1.5)flowPeaks(x,tol=0.1,h0=1,h=1.5)
x |
a data matrix for the flow cytometry data, it needs to have at least two rows, and the names for each column should be unique. For a flowFrame data, its exprssion matrix slot should be used as x, where only channles of interest are selected (see the example below). |
tol |
The tolerance (between 0 and 1) when neighboring clusters should be considered to be merged |
h0 |
The multiplier of the vaiarance matrix S0 |
h |
The multiplier of the variance matrix S |
It returns an object of class flowPeaks, which is a list of the following variables:
peaks.cluster |
An integer shows the cluster labels (between 1 and K for K clusters) for each cell. The clustering is based on the flowPeaks algorithm |
peaks |
A summary of the cluster information. It is a list with the following three variables:
|
kmeans.cluster |
An integer shows the cluster labels for the initial kmeans clustering |
kmeans |
A summary of the initial kmeans clustering. The meaning of the variables can be seens in the description of peaks above |
info |
The information that can be used for plot, and how the initial kmeans clustering and the final flowPeaks clustering are connected |
x |
The input data x |
Yongchao Ge [email protected]
Ge Y. et al, flowPeaks: a fast unsupervised clustering for flow cytometry data via K-means and density peak finding, 2012, Bioinformatics 8(15):2052-8
##demonstrate how to use a flowFrame ## Not run: require(flowCore) samp <- read.FCS(system.file("extdata","0877408774.B08", package="flowCore")) ##do the clustering based on the asinh transforamtion of ##the first two FL channels fp<-flowPeaks(asinh(samp@exprs[,3:4])) plot(fp) ## End(Not run) data(barcode) fp<-flowPeaks(barcode[,c(1,3)]) plot(fp) ##to compare it with the gold standard evalCluster(barcode.cid,fp$peaks.cluster,method="Vmeasure") #to remove the outliers fpc<-assign.flowPeaks(fp,fp$x) plot(fp,classlab=fpc,drawboundary=FALSE, drawvor=FALSE,drawkmeans=FALSE,drawlab=TRUE) #to adjust the cluster by increasing the tol,h0, h, which results #in a smaller number of clusters fp2<-adjust.flowPeaks(fp,tol=0.5,h0=2,h=2) summary(fp2) print(fp) #an alternative of using summary(fp)##demonstrate how to use a flowFrame ## Not run: require(flowCore) samp <- read.FCS(system.file("extdata","0877408774.B08", package="flowCore")) ##do the clustering based on the asinh transforamtion of ##the first two FL channels fp<-flowPeaks(asinh(samp@exprs[,3:4])) plot(fp) ## End(Not run) data(barcode) fp<-flowPeaks(barcode[,c(1,3)]) plot(fp) ##to compare it with the gold standard evalCluster(barcode.cid,fp$peaks.cluster,method="Vmeasure") #to remove the outliers fpc<-assign.flowPeaks(fp,fp$x) plot(fp,classlab=fpc,drawboundary=FALSE, drawvor=FALSE,drawkmeans=FALSE,drawlab=TRUE) #to adjust the cluster by increasing the tol,h0, h, which results #in a smaller number of clusters fp2<-adjust.flowPeaks(fp,tol=0.5,h0=2,h=2) summary(fp2) print(fp) #an alternative of using summary(fp)
This function takes the results generated from flowPeaks as an input, and plot the data in 2D. These plots display the clustering structure
## S3 method for class 'flowPeaks' plot(x,idx=c(1,2),drawlab=FALSE, cols=c("red","green3","blue","cyan","magenta","yellow","gray"),drawvor=TRUE, drawlocalpeaks=FALSE,drawkmeans=TRUE,drawboundary=TRUE, classlab, negcol, negpch,...)## S3 method for class 'flowPeaks' plot(x,idx=c(1,2),drawlab=FALSE, cols=c("red","green3","blue","cyan","magenta","yellow","gray"),drawvor=TRUE, drawlocalpeaks=FALSE,drawkmeans=TRUE,drawboundary=TRUE, classlab, negcol, negpch,...)
x |
Anobject of class flowPeaks, e.g., t
the output from the functions |
idx |
The index of the columns will be used to plot the clustering. idx needs to be at least legnth 2, and have no duplicate elements, and the values can only take from 1 to d, where d is the number of columns for the input matrix x that is used as an input of the function flowPeaks |
drawlab |
The option to decide whether we should draw the cluster labels |
cols |
The color specification for plotting the points in each cluster. Please note, "white" and "black" are not allowed, which are reserved for other purpse |
drawvor |
Deciding whether the voronoi diagram should be drawn, only good for 2D data |
drawlocalpeaks |
Decding whether the local peaks with a triangle symbol should be drawn |
drawkmeans |
Deciding whether the kmeans center with a filled circle should be drawn |
drawboundary |
Deciding whether the boudary between clusters should be drawn, only good for 2D data |
classlab |
Use this to replace the default class labels from
x$peak.cluster, for example, the classlab may come from
|
negcol |
Deciding the color of the negative, which are outliers |
negpch |
Deciing the symbols for the outliers |
... |
Optional additional arguments. At present no additional arguments are used. |
Yongchao Ge [email protected]
The display of the flowPeaks results
## S3 method for class 'flowPeaks' print(x,...)## S3 method for class 'flowPeaks' print(x,...)
x |
The output from the function |
... |
Optional additional arguments. At present no additional arguments are used. |
Yongchao Ge [email protected]
The summary of the flowPeaks results
## S3 method for class 'flowPeaks' summary(object,...)## S3 method for class 'flowPeaks' summary(object,...)
object |
The output from the function |
... |
Optional additional arguments. At present no additional arguments are used. |
Yongchao Ge [email protected]