Note: MCnebula2 is currently available, with Better compatibility and performance, please move to: https://github.com/Cao-lab-zcmu/MCnebula2.
MCnebula has been published at https://cao-lab-zcmu.github.io/MCnebula/.
Guidance for MCnebula application:
MCnebula_workflow.
The following installation has been tested in Pop!_OS 20.04 (Ubuntu 20.04).
We have tested MCnebula in R version 4.2 and 4.1. Maybe R version ≥ 4.0 is feasible. Herein, the codes are giving an example with installing R 4.2. In bash:
sudo apt-key adv \
--keyserver keyserver.ubuntu.com \
--recv-keys E298A3A825C0D65DFD57CBB651716619E084DAB9
sudo add-apt-repository \
"deb https://cloud.r-project.org/bin/linux/ubuntu $(lsb_release -cs)-cran40/"
sudo apt-get update
sudo apt install --no-install-recommends r-base
In bash:
## Libraries for installing 'usethis' and 'devtools'.
sudo apt install libssl-dev libcurl4-openssl-dev libblas-dev liblapack-dev libgfortran-11-dev
## Libraries for installing 'BiocManager' and its some packages.
sudo apt install libnetcdf-dev libopenbabel-dev libeigen3-dev
## Libraries For installing other graphic packages.
sudo apt install libfontconfig1-dev librsvg2-dev libmagick++-dev
Some data processing tools:
install.packages(c("data.table", "dplyr", "ggplot2",
"stringr", "tidyr", "reshape2",
"pbapply", "ggsci"))
Installing R packages from Bioconductor:
if (!require("BiocManager", quietly = TRUE))
install.packages("BiocManager")
BiocManager::install(c("MSnbase", "ChemmineOB"))
Of note, for current, 'ChemmineOB' is only available in Linux. Fortunately, 'ChemmineOB' is non-essential for MCnebula. However, the chemical structure mapping in child-nebula is depending on 'ChemmineOB'.
Installing graphic tools:
install.packages(c("igraph", "ggraph", "svglite",
"tidygraph", "rsvg", "grImport2",
"ggimage", "ggtext", "ggsci"))
For installing packages in github:
install.packages(c("usethis", "devtools"))
MCnebula is available in github:
devtools::install_github("Cao-lab-zcmu/MCnebula")
Loading library.
library(ggplot2)
library(ggraph)
library(grid)
library(MCnebula)
MCnebula perform data collating and integration in SIRIUS project.
First, user should initialize MCnebula at the directory of SIRIUS project.
path <- "my.sirius.project"
initialize_mcnebula(path)
The above will set some global var:
> .MCn.
.MCn.output .MCn.palette_label .MCn.palette_stat .MCn.sirius
.MCn.palette .MCn.palette_ppcp .MCn.results
Then, for data collating:
collate_structure(
exclude_element = c("Cl", "S", "P"),
ppm_error = 20
)
build_classes_tree_list()
collate_ppcp(
nebula_class = T,
## -------------------------------------
## detail control for nebula-class
## level 5, subclass, class, superclass
hierarchy_priority = c(6, 5, 4, 3),
max_classes = NA,
## -------------------------------------
## summarise nebula-index
nebula_index = T,
## -------------------------------------
## detail control for nebula-index
min_possess = 30,
max_possess_pct = 0.07,
filter_identical = c("top_hierarchy" = 4),
identical_factor = 0.7,
## following maybe remove lots of classes with many low score structures,
## if not, set NA
## ------------------
filter_via_struc_score = "tanimotoSimilarity",
struc_score_cutoff = 0.3,
min_reached_pct = 0.6,
## ------------------
rm_position_describe_class = T
## -------------------------------------
)
collate_ppcp is flexible with multiple augments. In addition, ignore the complexity, user can jsut run likewise:
collate_ppcp(
min_possess = 30,
max_possess_pct = 0.2
)
As well, the above commands will done with some global vars (data.frame or list) of which the name
begin with .MCn..
Third, for generate chemical nebulae (network):
generate_parent_nebula(
rm_parent_isolate_nodes = T
)
generate_child_nebulae(
max_edges = 5,
## this will write a output of .graphml, which support by Cytoscape.
output_format = "graphml"
)
Fourth, visualization for nebulae:
visualize_parent_nebula(
layout = "mds",
## map nodes color with superclass
nodes_color = c("hierarchy" = 3)
)
visualize_child_nebulae(
layout = "fr",
## herein, you can mark some features with specific color.
## if not, just ignore the augment.
nodes_mark = data.frame(
.id = c("1", "300", "500"),
mark = c("biomarker1", "biomarker2", "biomarker3")
),
palette = c(
biomarker1 = 'pink',
biomarker2 = 'lightblue',
biomarker3 = 'white',
Others = '#B8B8B8'
),
nodes_size_range = c(2, 4),
## adjust the legend
remove_legend_lab = T,
legend_fill = T,
legend_position = "bottom",
nodes_stroke = 0,
edges_color = "#D9D9D9"
)
Last, users are encouraged for in-depth visualization of child-nebula:
annotate_child_nebulae(
## string, i.e. class name in nebula-index
nebula_name,
layout = "fr",
## a table to mark color of nodes
nodes_mark = data.frame(
.id = c("1", "300", "500"),
mark = c("biomarker1", "biomarker2", "biomarker3")
),
plot_nodes_id = T,
plot_structure = T,
plot_ppcp = T,
## manually define the color of nodes
palette = c(
biomarker1 = 'pink',
biomarker2 = 'lightblue',
biomarker3 = 'white',
Others = '#B8B8B8'
),
## feature quantification table
ratio_df = data.frame(
.id = 1:1000,
sample1 = rnorm(1000, 1E5, 1E4),
sample2 = rnorm(1000, 1E6, 5E4),
sample3 = rnorm(1000, 1E4, 5E3),
),
## A vector of the hex color with names or not
palette_stat = c(
sample1 = 'blue',
sample2 = 'yellow',
sample3 = 'red'
),
## control nodes size in child-nebula, zoom in or zoom out globally.
global.node.size = 0.8,
## the args of `ggplot::theme`
theme_args = list(
panel.background = element_rect(),
panel.grid = element_line()
),
return_plot = F
)
...
Use InChiKey planar to get all possible InChIKey.
pubchem_curl_inchikey function use RCurl package and through PubChem API to get InChIKey.
We design this function with the stability of the network in consideration.
Therefore, the process of getting data will not fail due to Status: 503.
pubchem_curl_inchikey(
## a vector of InChIKey planar
inchikey2d,
## output path
dir,
## number of threads
curl_cl = NULL,
## each InChiKey planar will get a .TSV file. Here, this setting could gather them
## as 'list' into .rdata. Indeed, this can save storage space.
gather_as_rdata = T
)
## if 'gather_as_rdata' == T, user can use following function to get results.
inchikey_set <- extract_rdata_list(
paste0(dir, "/", "inchikey.rdata")
)
Use PubChem CID to get synonyms of compounds.
Note that pubchem_curl_inchikey not only get InChIKey, but also PubChem CID.
Herein, pubchem_get_synonyms is used to get synonyms via PubChem CID.
pubchem_get_synonyms(
## a vector of PubChem CID
cid,
## output path
dir,
## number of threads
curl_cl = NULL,
gather_as_rdata = T
)
syno_set <- extract_rdata_list(
paste0(dir, "/", "cid.rdata")
)