1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
|
% Generated by roxygen2: do not edit by hand
% Please edit documentation in R/Optimum_KernelC.R
\name{Optimum_KernelC}
\alias{Optimum_KernelC}
\title{Kernel function for optimizing parameters and hidden variables in DeMixT}
\usage{
Optimum_KernelC(
inputdata,
groupid,
nspikein,
setting.pi,
givenpi,
givenpiT,
niter,
ninteg,
tol,
sg0 = 0.5^2,
mu0 = 0,
pi01 = NULL,
pi02 = NULL,
nthread = 1
)
}
\arguments{
\item{inputdata}{A matrix of expression data (e.g gene expressions) from
reference (e.g. normal) and mixed samples (e.g. mixed tumor samples). It is
a \eqn{G*M} matrix where \eqn{G} is the number of genes and \eqn{M} is the
number of samples including reference and mixed samples. Samples with the
same tissue type should be placed together in columns (e.g. cbind(normal
amples, mixed tumor samples).}
\item{groupid}{A vector of indicators to denote if the corresponding samples
are reference samples or mixed tumor samples. DeMixT is able to deconvolve
mixed tumor samples with at most three components. We use 1 and 2 to denote
the samples referencing the first and the second known component in mixed
tumor samples. We use 3 to indicate mixed tumor samples prepared to be
deconvolved. For example, in two-component deconvolution, we have
c(1,1,...,3,3) and in three-component deconvolution, we have
c(1,1,...,2,2,...,3,3).}
\item{nspikein}{The number of spikes in normal reference used for proportion
estimation. The default value is \eqn{ min(200, 0.3*My)}, where \eqn{My} the
number of mixed tumor samples. If it is set to 0, proportion estimation is
performed without any spike in normal reference.}
\item{setting.pi}{If it is set to 0, then deconvolution is performed without
any given proportions; if set to 1, deconvolution with given proportions for
the first and the second known component is run; if set to 2, deconvolution
is run with given tumor proportions. This option helps to perform
deconvolution in different settings. In estimation of component-specific
proportions, we use a subset of genes ; so when it is required to deconvolve
another subset of genes, we just easily plug back our estimated proportions
by setting this option to 1. In our two-step estimation strategy in a
three-component setting, this option is set to 2 to implement the second
step.}
\item{givenpi}{\eqn{ST}-Vector of proportions. Given the number of mixed
tumor samples is \eqn{My(My<M)}, \eqn{ST} is set to \eqn{2*My} in a
three-component setting and \eqn{My} in a two-component setting. When
setting.pi is 1, it is fixed with the given proportions for the first and
the second known component of mixed tumor samples, or for one unknown
component when there is just one type of reference tissues. It has the form
of Vector \eqn{PiN1-1}, \eqn{PiN1-2}, ..., \eqn{PiN1-My}, \eqn{PiN2-1},
\eqn{PiN2-2}, ..., \eqn{PiN2-My}.}
\item{givenpiT}{\eqn{ST}-Vector of proportions. When setting.pi is set to 2,
givenpiT is fixed with given proportions for unknown component of mixed
tumor samples. This option is used when we adopt a two-step estimation
strategy in deconvolution. It has the form of Vector \eqn{PiT-1},
\eqn{PiT-2}, ..., \eqn{PiT-My}. If option is not 2, this vector can be given
with any element.}
\item{niter}{The number of iterations used in the algorithm of iterated
conditional modes. A larger value can better guarantee the convergence in
estimation but increase the computation time.}
\item{ninteg}{The number of bins used in numerical integration for computing
complete likelihood. A larger value can increase accuracy in estimation but
also increase the running time. Especially in three-component deconvolution,
the increase of number of bins can greatly lengthen the running time.}
\item{tol}{The convergence criterion. The default is 10^(-5).}
\item{sg0}{Initial value for \eqn{\sigma^2}. The default is 0.5^2.}
\item{mu0}{Initial value for \eqn{\mu}. The default is 0.}
\item{pi01}{Initialized proportion for first kown component. The default is
\eqn{Null} and pi01 will be generated randomly from uniform distribution.}
\item{pi02}{Initialized proportion for second kown component. pi02 is needed
only for running a three-component model. The default is \eqn{Null} and pi02
will be generated randomly from uniform distribution.}
\item{nthread}{The number of threads used for deconvolution when OpenMP is
available in the system. The default is the number of whole threads minus
one. In our no-OpenMP version, it is set to 1.}
}
\value{
\item{pi}{Matrix of estimated proportions for each known component.
The first row corresponds to the proportion estimate of each sample for the
first known component (groupid = 1) and the second row corresponds to that
for the second known component (groupid = 2).} \item{decovExpr}{A matrix of
deconvolved expression profiles corresponding to unknown (e.g tumor)
component in mixed samples for a given subset of genes. Each row corresponds
to one gene and each column corresponds to one sample.}
\item{decovMu}{Estimated \eqn{Mu} of log2-normal distribution for tumor
component.} \item{decovSigma}{Estimated \eqn{Sigma} of log2-normal
distribution for tumor component.} \item{pi1}{An \eqn{My*I} matrix of
estimated proportions for each iteration, where \eqn{I} is the number of
iteration, for the first known component.} \item{pi2}{An \eqn{My*I} matrix
of estimated proportions for each iteration, where \eqn{I} is the number of
iteration, for the second known component.}
}
\description{
This function is invoked by DeMixT_GS or DeMixT_DE and DeMixT_S2 to finish
parameter estimation by iterated conditional mode algorithm and reconstitute
gene expression profile of all components.
}
\examples{
# Example 1: simulated two-component data
data(test.data.2comp)
# data.N1 <- SummarizedExperiment::assays(test.data.2comp$data.N1)[[1]]
# data.Y <- SummarizedExperiment::assays(test.data.2comp$data.Y)[[1]]
# inputdata <- cbind(data.N1, data.Y)
# groupid <- c(rep(1, ncol(data.N1)), rep(3, ncol(data.Y)))
# nspikein <- 0
# Optimum_KernelC(inputdata, groupid,
# nspikein = nspikein, setting.pi = 0,
# givenpi = rep(0, 2 * ncol(data.y)),
# niter = 10, ninteg = 30, tol = 10^(-4))
}
\references{
Wang Z, Cao S, Morris J S, et al. Transcriptome Deconvolution of
Heterogeneous Tumor Samples with Immune Infiltration. iScience, 2018, 9:
451-460.
}
\seealso{
http://bioinformatics.mdanderson.org/main/DeMixT
}
\author{
Zeya Wang, Wenyi Wang
}
\keyword{Optimum_KernelC}
|
