Super quick start

To quickly demonstrate the application of the ‘MCnebula2’ R package, we will then use a data extracted from a finished SIRIUS project (using the collate_used function) to demonstrate its use. The following demonstration can be repeated in R as long as the user has installed the MCnebula2 R package (and possibly some small R package, such as RCurl).

Set-up #

library(MCnebula2)
if (!requireNamespace("RCurl")) {
  install.packages("RCurl")
}

Download and load the demo data #

Note: The following object (toActiv30) just contain 30 Features. However, the real untargeted LC-MS/MS dataset could hardly be such a small amount of Features. We collected this dataset only to demonstrate the application of the MCnebula2 R package, please do not apply the parameters involved below in real data analysis (for application of the parameters, please refer to workflow: Metabolic workflow; Chemcal workflow; or use help() with your confused functions or methods).

rda <- paste0(tmp <- tempdir(), "/toActiv30.rdata")
bin <- RCurl::getURLContent("https://raw.githubusercontent.com/Cao-lab-zcmu/utils_tool/master/inst/extdata/toActiv30.rdata")
writeBin(bin, rda)

load(rda)
mcn <- toActiv30

Filter candidates and integrate data #

All basic workflow was run as following. Again, the parameters used in following was just with the purpose of demonstration, do not applied these in real analysis.

mcn <- filter_structure(mcn)
mcn <- create_reference(mcn)
mcn <- filter_formula(mcn, by_reference = T)

mcn <- create_stardust_classes(mcn)
mcn <- create_features_annotation(mcn)
mcn <- cross_filter_stardust(mcn, 5, .7)

mcn <- create_nebula_index(mcn)
mcn <- compute_spectral_similarity(mcn)
mcn <- create_parent_nebula(mcn, 0.01, 5, T)
mcn <- create_child_nebulae(mcn, 0.01, 5)

mcn <- create_parent_layout(mcn)
mcn <- create_child_layouts(mcn)
mcn <- activate_nebulae(mcn)

Visualize Parent-Nebula and Child-Nebula #

Drawing these Nebula and saved as .svg.

Parent-Nebula #

For Parent-Nebula:

p <- visualize(mcn, 'parent')
ggsave("parent_nebula.svg", p, height = 4)

Child-Nebulae #

For Child-Nebulae:

svg("child_nebulae.svg", 9, 9)
visualize_all(mcn)
dev.off()

Quantification analysis #

These was no quantification data in object mcn. So we need to simulate a quantification dataset into that. A predefined function could done this well:

mcn <- MCnebula2:::.simulate_quant_set(mcn)

Let’s check what we have simulated:

sample_metadata(mcn)

## # A tibble: 20 × 2
##    sample    group  
##    <chr>     <chr>  
##  1 control_1 control
##  2 control_2 control
##  3 control_3 control
##  4 control_4 control
##  5 control_5 control
##  6 model_1   model  
##  7 model_2   model  
##  8 model_3   model  
##  9 model_4   model  
## 10 model_5   model  
## 11 treat_1   treat  
## 12 treat_2   treat  
## 13 treat_3   treat  
## 14 treat_4   treat  
## 15 treat_5   treat  
## 16 pos_1     pos    
## 17 pos_2     pos    
## 18 pos_3     pos    
## 19 pos_4     pos    
## 20 pos_5     pos

features_quantification(mcn)

## # A tibble: 30 × 21
##    .features_id control_1 control_2 control_3 control_4 control_5 model_1
##    <chr>            <dbl>     <dbl>     <dbl>     <dbl>     <dbl>   <dbl>
##  1 202               29.1      57.4      18.5      60.5      46.8    27.5
##  2 2020              32.9      63.0      56.6      63.9      69.9    66.1
##  3 2022              73.8      23.6      31.9      45.5      41.5    29.5
##  4 2023              42.6      70.2      71.8      39.1      10.7    64.4
##  5 2024              43.9      78.6      77.4      75.6      58.7    34.8
##  6 2025              24.5      37.2      58.7      25.7      15.6    48.1
##  7 2026              89.8      48.4      52.3      68.2      24.3    57.3
##  8 2027              80.5      66.7      62.1      52.5      48.3    52.2
##  9 2028              55.3      91.3      73.6      76.4      74.1    28.7
## 10 2029              56.6      30.9      15.6      34.6      42.6    21.1
## # … with 20 more rows, and 14 more variables: model_2 <dbl>, model_3 <dbl>,
## #   model_4 <dbl>, model_5 <dbl>, treat_1 <dbl>, treat_2 <dbl>, treat_3 <dbl>,
## #   treat_4 <dbl>, treat_5 <dbl>, pos_1 <dbl>, pos_2 <dbl>, pos_3 <dbl>,
## #   pos_4 <dbl>, pos_5 <dbl>

Statistic analysis #

Compare two groups: Model vs Control.

mcn <- binary_comparison(mcn, model - control)
top_data <- top_table(statistic_set(mcn))[[ 1 ]]
top_data

## # A tibble: 30 × 7
##    .features_id  logFC AveExpr     t  P.Value adj.P.Val      B
##    <chr>         <dbl>   <dbl> <dbl>    <dbl>     <dbl>  <dbl>
##  1 2071         -1.76  -0.183  -3.75 0.000630    0.0162 -0.426
##  2 2025          1.22   0.0974  3.56 0.00108     0.0162 -0.919
##  3 2097          1.02  -0.0853  2.85 0.00731     0.0731 -2.66 
##  4 2096         -0.993 -0.0310 -2.56 0.0149      0.112  -3.29 
##  5 209           0.900 -0.0986  2.42 0.0209      0.126  -3.59 
##  6 2095         -0.704 -0.104  -1.82 0.0773      0.386  -4.69 
##  7 2092          0.894 -0.189   1.50 0.142       0.610  -5.18 
##  8 2023          0.602  0.120   1.34 0.188       0.658  -5.39 
##  9 2099          0.575 -0.126   1.25 0.221       0.658  -5.50 
## 10 2029          0.474 -0.0430  1.19 0.242       0.658  -5.57 
## # … with 20 more rows

Tracing top features in Child-Nebulae #

Show Child-Nebulae as tracing mode to find where the top Features were classified.

## Here we just set top 5 to trace
tops <- top_data$.features_id[1:5]
mcn2 <- set_tracer(mcn, tops)

mcn2 <- create_child_nebulae(mcn2)
mcn2 <- create_child_layouts(mcn2)
mcn2 <- activate_nebulae(mcn2)
mcn2 <- set_nodes_color(mcn2, use_tracer = T)

svg("tracer_child_nebulae.svg")
visualize_all(mcn2)
dev.off()

Others #

More demonstration can be found in workflow