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
|
% Generated by roxygen2: do not edit by hand
% Please edit documentation in R/determine_regional_similarity.R
\name{determine_regional_similarity}
\alias{determine_regional_similarity}
\title{Determine regional mutation pattern similarity}
\usage{
determine_regional_similarity(
vcf,
ref_genome,
chromosomes,
window_size = 100,
stepsize = 25,
extension = 1,
oligo_correction = FALSE,
exclude_self_mut_mat = TRUE,
max_window_size_gen = 2e+07,
verbose = FALSE
)
}
\arguments{
\item{vcf}{GRanges object}
\item{ref_genome}{BSgenome reference genome object}
\item{chromosomes}{Vector of chromosome/contig names of the reference genome
to be plotted.}
\item{window_size}{The number of mutations in a window. (Default: 100)}
\item{stepsize}{The number of mutations that a window slides in each step.
(Default: 25)}
\item{extension}{The number of bases, that's extracted upstream and
downstream of the base substitutions, to create the mutation matrices.
(Default: 1).}
\item{oligo_correction}{Boolean describing whether oligonucleotide frequency
correction should be applied. (Default: FALSE)}
\item{exclude_self_mut_mat}{Boolean describing whether the mutations in a
window should be subtracted from the global mutation matrix. (Default:
TRUE)}
\item{max_window_size_gen}{The maximum size of a window before it is removed.
(Default: 20,000,000)}
\item{verbose}{Boolean determining the verbosity of the function. (Default: FALSE)}
}
\value{
A "region_cossim" object containing both the cosine similarities and
the settings used in this analysis.
}
\description{
Calculate the cosine similarities between the global mutation profile and the
mutation profile of smaller genomic windows, using a sliding window approach.
Regions with a very different mutation profile can be identified in this way.
This function generally requires many mutations (~100,000) to work properly.
}
\details{
First a global mutation matrix is calculated using all mutations.
Next, a sliding window is used. This means that we create a window
containing the first x mutations. The cosine similarity, between the
mutation profiles of this window and the global mutation matrix, is then
calculated. The window then slides y mutations to the right and the cosine
similarity is again calculated. This process is repeated until the final
mutation on a chromosome is reached. This process is performed separately
per chromosome. Windows that span a too large region of the genome are
removed, because they are unlikely to contain biologically relevant
information.
The number of mutations that the window slides to the right in each step is
called the stepsize. The best stepsize depends on the window size. In
general, we recommend setting the stepsize between 25% and 100% of the
window size.
The analysis can be performed for trinucleotides contexts, for a larger
context, or for just the base substitutions. A smaller context might miss
detailed differences in mutation profiles, but is also less noisy. We
recommend using a smaller extension when analyzing small datasets.
It's possible to correct for the oligonucleotide frequency of the windows.
This is done by calculating the cosine similarity of the oligonucleotide
frequency between each window and the genome. The cosine similarity of the
mutation profiles is then divided by the oligonucleotide similarity. This
ensures that regions with an abnormal oligonucleotide frequency don't show
up as having a very different profile. The oligonucleotide frequency
correction slows down the function, so we advise the user to keep it off
for exploratory analyses and to only turn it on to validate interesting
results.
By default the mutations in a window are subtracted from the global
mutation matrix, before calculating the cosine similarity. This increases
sensitivity, but could also decrease specificity. This subtraction can be
turned of with the 'exclude_self_mut_mat' argument.
}
\examples{
## See the 'read_vcfs_as_granges()' example for how we obtained the
## following data:
grl <- readRDS(system.file("states/read_vcfs_as_granges_output.rds",
package = "MutationalPatterns"
))
## We pool all the variants together, because the function doesn't work well
## with a limited number of mutations. Still, in practice we recommend to use
## more mutations that in this example.
gr = unlist(grl)
## Specifiy the chromosomes of interest.
chromosomes <- names(genome(gr)[1:3])
## Load the corresponding reference genome.
ref_genome <- "BSgenome.Hsapiens.UCSC.hg19"
library(ref_genome, character.only = TRUE)
## Determine the regional similarities. Here we use a small window size to make the function work.
## In practice, we recommend a larger window size.
regional_sims = determine_regional_similarity(gr,
ref_genome,
chromosomes,
window_size = 40,
stepsize = 10,
max_window_size_gen = 40000000
)
## Here we use an extensiof of 0 to reduce noise.
## We also turned verbosity on, so you can see at what step the function is.
## This can be useful on large datasets.
regional_sims_0_extension = determine_regional_similarity(gr,
ref_genome,
chromosomes,
window_size = 40,
stepsize = 10,
extension = 0,
max_window_size_gen = 40000000,
verbose = TRUE
)
}
\seealso{
\code{\link{plot_regional_similarity}}
Other regional_similarity:
\code{\link{plot_regional_similarity}()}
}
\concept{regional_similarity}
|
