| Title: | Dirichlet Process Beta-Binomial Mixture |
|---|---|
| Description: | Beta-binomial Mixture Model is used to infer the pattern from count data. It can be used for clustering of RNA methylation sequencing data. |
| Authors: | Lin Zhang |
| Maintainer: | Lin Zhang <[email protected]> |
| License: | GPL (>= 2) |
| Version: | 0.2.5 |
| Built: | 2024-11-02 04:11:54 UTC |
| Source: | https://github.com/cran/DPBBM |
This package is developed for the beta-binomial mixture model based clustering
| Package: | DPBBM |
| Type: | Package |
| Version: | 1.0.1 |
| Date: | 2016-06-26 |
| License: | GPL-2 |
Coming soon!
# Please check the main function of the package ?dpbbm_mc_iterations# Please check the main function of the package ?dpbbm_mc_iterations
This is to generate the simulation data based on Beta-bionomial mixture model
bbm_data_generate(S=3, G=50, K=3, prob=rep(1,times=3), alpha_band=c(2,6), beta_band=c(2,6), nb_mu=100,nb_size=0.2, plotf = FALSE, max_cor=0.5)bbm_data_generate(S=3, G=50, K=3, prob=rep(1,times=3), alpha_band=c(2,6), beta_band=c(2,6), nb_mu=100,nb_size=0.2, plotf = FALSE, max_cor=0.5)
S |
Number of samples in the simulated data |
G |
Number of sites in the simulated data |
K |
Number of clusters that exist in the simulated data |
prob |
the cluster weight for each cluster |
alpha_band |
the region used to generate the parameter of beta distribution alpha |
beta_band |
the region used to generate the parameter of beta distribution beta |
nb_mu |
alternative parametrization via mean for Negative Binomial distribution |
nb_size |
target for number of successful trials, or dispersion parameter (the shape parameter of the gamma mixing distribution) for Negative binomial distrition. Must be strictly positive, need not be integer. |
plotf |
option for whether plot the generated data according to clusters or not |
max_cor |
The maximized correlation allowed for the simulated data, which used to guarantee the data is not highly correlated. |
The Dirichlet Process based beta-binomial mixture model clustering
The function returns simulation data generated based on beta binomial mixture model
Lin Zhang, PhD <[email protected]>
Reference coming soon!
set.seed(123455) S <- 4 G <- 100 K <- 3 nb_mu <- 100 nb_size <- 0.8 prob <- c(1,1,1) mat <- bbm_data_generate(S=S,G=G,K=K,prob=prob,alpha_band=c(2,6),beta_band=c(2,6), nb_mu=nb_mu,nb_size=nb_size, plotf = TRUE, max_cor=0.5) table(mat$gamma) pie(mat$gamma) id <- order(mat$gamma); c <- mat$gamma[id] mat_ratio <- (mat$k+1)/(mat$n+1); heatmap(mat_ratio[id,], Rowv = NA, Colv = NA, scale="none", RowSideColors=as.character(c), xlab = "4 samples", ylab="100 RNA methylation sites")set.seed(123455) S <- 4 G <- 100 K <- 3 nb_mu <- 100 nb_size <- 0.8 prob <- c(1,1,1) mat <- bbm_data_generate(S=S,G=G,K=K,prob=prob,alpha_band=c(2,6),beta_band=c(2,6), nb_mu=nb_mu,nb_size=nb_size, plotf = TRUE, max_cor=0.5) table(mat$gamma) pie(mat$gamma) id <- order(mat$gamma); c <- mat$gamma[id] mat_ratio <- (mat$k+1)/(mat$n+1); heatmap(mat_ratio[id,], Rowv = NA, Colv = NA, scale="none", RowSideColors=as.character(c), xlab = "4 samples", ylab="100 RNA methylation sites")
This is to evaluate the clustering method with Bcubed F sore.
BCubed_metric(L, C, alpha)BCubed_metric(L, C, alpha)
L |
real label of classes |
C |
classification label of classes drawn by clustering method |
alpha |
F metric parameter which used to average precision and recall |
The clustering evaluation method based on Bcubed F metric.
The function returns Bcubed F score of the clustering method. The higher the value is, the better performance the clustering method can get.
Lin Zhang, PhD <[email protected]>
Reference coming soon!
L <- c(1,1,2,1,1,2,2) C <- c(2,2,1,2,2,1,1) alpha <- 0.5 Bcubed_score <- BCubed_metric(L, C, alpha) Bcubed_scoreL <- c(1,1,2,1,1,2,2) C <- c(2,2,1,2,2,1,1) alpha <- 0.5 Bcubed_score <- BCubed_metric(L, C, alpha) Bcubed_score
This is the Markov Chain Monte Carlo iterations for DPBBM
dpbbm_mc_iterations(x, size.x, m = 1, max_iter = 2000, a = 0.1, b = 1, tau = 1, sig_alpha = 25/9, sig_beta = 25/9, tau.method = "auto", debug = FALSE)dpbbm_mc_iterations(x, size.x, m = 1, max_iter = 2000, a = 0.1, b = 1, tau = 1, sig_alpha = 25/9, sig_beta = 25/9, tau.method = "auto", debug = FALSE)
x |
a matrix of k for clustering, referring to IP reads in m6A seq data |
size.x |
a matrix of n for clustering, referring to the summation of IP reads and input reads in m6A seq data |
m |
a value indicating the auxiliary clusters used in DPBBM |
max_iter |
maximized iterations in DPBBM |
a |
Hyperparameter a for tau |
b |
Hyperparameter b for tau |
tau |
Prior for tau |
sig_alpha |
variation for parameter alpha of beta distribution |
sig_beta |
variation for parameter beta of beta distribution |
tau.method |
tau.method should be set to "auto" or "stable", refer to tau for detail description. |
debug |
whether DPBBM print the debug info or not. Default: FALSE |
The Dirichlet Process based beta-binomial mixture model clustering
The function returns the cluster label withdrawn by DPBBM
Lin Zhang, PhD <[email protected]>
Reference coming soon!
# generate a simulated dataset set.seed(123455) S <- 4 G <- 100 K <- 3 nb_mu <- 100 nb_size <- 0.8 prob <- c(1,1,1) mat <- bbm_data_generate(S=S,G=G,K=K,prob=prob,alpha_band=c(2,6),beta_band=c(2,6), nb_mu=nb_mu,nb_size=nb_size, plotf = FALSE, max_cor=0.5) # check generated data id <- order(mat$gamma); c <- mat$gamma[id] mat_ratio <- (mat$k+1)/(mat$n+1); heatmap(mat_ratio[id,], Rowv = NA, Colv = NA, scale="none", RowSideColors=as.character(c), xlab = "4 samples", ylab="100 RNA methylation sites") ## Run the DPBBM result. This step takes a really long time. ## You are suggested to check the pre-prepared example for a quick start F=system.file("extdata", "DPBBM_example.html", package="DPBBM") browseURL(url=F) ## Alternatively # cluster_label <- dpbbm_mc_iterations(mat$k, mat$n) # # Show the clustering result. # table(cluster_label) # pie(table(mat$gamma)) # # # Compare the clustering result with the true clustering IDs. # id <- order(mat$gamma); # c <- cluster_label # r <- rainbow(3, start = 0, end = 0.3) # mat_ratio <- (mat$k+1)/(mat$n+1); # heatmap(mat_ratio[id,], Rowv = NA, Colv = NA, scale="none", # RowSideColors = as.character(cluster_label[id]), # margins = c(3,25))# generate a simulated dataset set.seed(123455) S <- 4 G <- 100 K <- 3 nb_mu <- 100 nb_size <- 0.8 prob <- c(1,1,1) mat <- bbm_data_generate(S=S,G=G,K=K,prob=prob,alpha_band=c(2,6),beta_band=c(2,6), nb_mu=nb_mu,nb_size=nb_size, plotf = FALSE, max_cor=0.5) # check generated data id <- order(mat$gamma); c <- mat$gamma[id] mat_ratio <- (mat$k+1)/(mat$n+1); heatmap(mat_ratio[id,], Rowv = NA, Colv = NA, scale="none", RowSideColors=as.character(c), xlab = "4 samples", ylab="100 RNA methylation sites") ## Run the DPBBM result. This step takes a really long time. ## You are suggested to check the pre-prepared example for a quick start F=system.file("extdata", "DPBBM_example.html", package="DPBBM") browseURL(url=F) ## Alternatively # cluster_label <- dpbbm_mc_iterations(mat$k, mat$n) # # Show the clustering result. # table(cluster_label) # pie(table(mat$gamma)) # # # Compare the clustering result with the true clustering IDs. # id <- order(mat$gamma); # c <- cluster_label # r <- rainbow(3, start = 0, end = 0.3) # mat_ratio <- (mat$k+1)/(mat$n+1); # heatmap(mat_ratio[id,], Rowv = NA, Colv = NA, scale="none", # RowSideColors = as.character(cluster_label[id]), # margins = c(3,25))