cells2numbers / migrationminer

MigrationmineR is a R package to analyse and profile in vitro cell tracking and migration data. It is belongs to the cytominer-verse used for morphological profiling and allows to create temporal or dynamic profiles.

migrationminer

migrationminer is a R package to analyse and profile in vitro cell tracking and migration data. It is belongs to the cytominer-verse used for morphological profiling and allows to create temporal or dynamic profiles.

It works well together with CellProiler and the CellProfiler tracking module. Beside that it is quite easy to parse other formats as long as the tracking data is available in tidy format. That means, that the data is available as csv where each row represents a cell with at least four columns

• x coordinate
• y coordinate
• time point
• track label

Installation

migrationminer is not yet available from CRAN but you can install directly from github. First, you need to install the devtools package. You can do this from CRAN. Invoke R and then type

install.packages("devtools")

Load the package using

library(devtools)

Now install migrationminer from github using

install_github("cells2numbers/migrationminer")

You should see something like

Downloading GitHub repo cells2numbers/migrationminer@master
from URL https://api.github.com/repos/cells2numbers/migrationminer/zipball/master
Installing migrationminer
'/usr/local/Cellar/r/3.4.1_2/lib/R/bin/R' --no-site-file --no-environ --no-save --no-restore --quiet CMD INSTALL  \
  '/private/var/folders/qq/v5mwx8g15655gt708s86vys8yzcpll/T/RtmpUyrzzK/devtoolse7f945c7b360/cells2numbers-migrationminer-8a95c5d'  \
  --library='/usr/local/lib/R/3.4/site-library' --install-tests 

* installing *source* package ‘migrationminer’ ...
** R
** tests
** preparing package for lazy loading
** help
*** installing help indices
** building package indices
** testing if installed package can be loaded
* DONE (migrationminer)
Reloading installed migrationminer

If this does not work for you or you find any bugs, please file an issue!

migrationminer's Issues

Add mean position for each track

replace data_frame() with tibble()

rename neutrominer to migrationminer

add function to correct for confounding variables

example code:

Correct for three variables (age, sex and date of recording). Each combination of subset / offset
is corrected separetly

# helper function to iterate over all features 
get_lm_helper_function <- function(profiles_tmp) {
  # get feature names 
  # feature_names <- colnames(profiles_tmp) %>% str_subset("^Track")
  feature_names <- c("Track_Angle", "Track_Directionality", "Track_CI","Track_Speed","Track_yFMI", "Track_xFMI", "Track_CI","Track_DP", "Track_MSD", "Track_Speed_X","Track_Speed_Y", "Track_Speed_max")
  # calculate corrected values as residuals of lm
  df_result <- do.call(
    cbind,
    lapply(
      feature_names, 
      get_lm_model_fitted_values, 
      profiles_tmp = profiles_tmp
      )
    )
  # add experiment id Metadata_experiment
  df_result %<>% 
    mutate(Metadata_experiment = profiles_tmp$Metadata_experiment)
}

# linear model lm is used to remove date/age/sex effects
get_lm_model_fitted_values <- function(y, profiles_tmp){
  new_colname <- paste0(y)
  
  tryCatch({
      lm_formula <- paste0(y, " ~ Clinical_Parameter_Age + Clinical_Parameter_Sex + Clinical_Parameter_Date")
      lm_model <- lm(lm_formula, data = profiles_tmp,na.action = na.exclude) 
      fitted_values <- tibble( "V1" = residuals(lm_model, na.action=na.exclude))
      colnames(fitted_values) <- new_colname
    },
    fitted_values = function(err){
      fitted_values <- tibble("V1" = rep(NA, nrow(profiles_tmp))) 
      colnames(fitted_values) <- new_colname
      return(fitted_values)
    }
  )

  return(fitted_values)
} 

Correct values

profiles_corrected <- profiles %>% 
  filter(!is.na(Track_Directionality), Metadata_Frame_Subset < 180, Metadata_offset < 180) %>% 
  group_by(Metadata_Frame_Subset, Metadata_offset) %>% 
  do(get_lm_helper_function(.))

Travis fails with missing libgit2

Libgit2 is missing on the build system,

adding libgit2 as a binary package as suggested here does not solve the issue:

https://stackoverflow.com/questions/53942148/travis-ci-cannot-find-packages-in-the-pkg-config-search-path

plot_windrose shows wrong x direction

Sample code

library(tidyverse)
library(migrationminer)

x = c(1,100,-2)
y = c(1,-200,1)

Track_Angle =
  atan2(tail(y, n = 1) - (y)[1],
        tail(x, n = 1) - (y)[1]
  ) %>% print

(Track_Angle >= pi / 4) &  (Track_Angle < 3 * pi / 4 )

# test windroseplot 


# up 1.5 
# down -1.5
# left 3 
# right 0

tracks <- tibble(Track_Angle = -c(-1.5, 3, -1.5, 3), Track_Speed = c(1,1,1,1), id = c(1,1,1,1))

plot_windrose(tracks)

add code coverage

replace ```.dots =``` using tidyeval

I would be nice to remove the call

group_by_(.dots = strata))

by

group_by(!!!strata)

using quasiequations. A good example / summary can be found here http://blog.jalsalam.com/posts/2017/quasi-quotation-applications/

add visualization

Add visualization including

• roseplot / windrose plot
• trajectory plot 2d/3d

Build fails because of wrong dependencies

Travis build fails because cytominer can not be installed via cran.

Add windowed profiles

Add functionality to calculate the migration parameter on a windowed frame. This allows quantifying the change of migration over time.

Sample code:

create_windowed_migration_profiles <- function(df_temp, t1, offset = 20) {
  df_temp %<>% 
    filter(Metadata_frame > t1) %>% 
    filter(Metadata_frame < (t1 + offset) ) 
  
  tracks <- migrationminer::track(
      df_temp, 
      c("TrackObjects_Label"),
      t_var = "Metadata_frame"
    )  %>% 
    filter(complete.cases(.)) %>% 
    filter(Track_Length > offset/2) %>% 
    summarise_all( mean) %>% 
    mutate(Metadata_Frame_Subset = t1) %>% 
    mutate(Metadata_offset = offset) 
    
  return(tracks)
} 

Temporal change of the migration parameter 'speed' over time. Different window sizes (offsets) were chosen. The shown data belongs to an experiment that is characterized by a strong flow

test fail with with 'unknown function "n"

create vignette and explain sector analysis

rewrite functions using tidyeval

dplyr 0.7 introduces tidy evaluation that should be used to rewrite the track functions, see https://blog.rstudio.com/2017/06/13/dplyr-0-7-0/

add pkgdown

Integrate pkgdown, see https://github.com/hadley/pkgdown