--- title: "Sojourner: an R package for statistical analysis of single molecule trajectories" author: - name: Sheng Liu affiliation: - Department of Biology, Krieger School of Arts and Sciences, Johns Hopkins University - name: Sun Jay Yoo affiliation: - Department of Biomedical Engineering, Johns Hopkins University - Department of Computer Science, Johns Hopkins University - name: Xiaona Tang affiliation: - Department of Biology, Krieger School of Arts and Sciences, Johns Hopkins University - name: Young Soo Sung affiliation: - Department of Computer Science, Johns Hopkins University - name: Carl Wu affiliation: - Department of Molecular Biology and Genetics, School of Medicine, Johns Hopkins University - Department of Biology, Krieger School of Arts and Sciences, Johns Hopkins University package: sojourner date: "`r Sys.Date()`" output: # rmarkdown::html_vignette: BiocStyle::html_document: toc: true toc_depth: 3 toc_float: true number_sections: true # output: # BiocStyle::pdf_document vignette: > %\VignetteIndexEntry{Sojourner: an R package for statistical analysis of single molecule trajectories} %\VignetteEngine{knitr::rmarkdown} %\VignetteEncoding{UTF-8} --- ```{r style, echo = FALSE, results = 'asis'} BiocStyle::markdown() ``` ![](sojourner_logo.png){width=30%} # Introduction A basic task in 2D single molecule tracking is to determine diffusion coefficient from molecule trajectories identified after initial image data acquisition. The trajectory (term used interchangeably with "tracks") of a molecule is presented as a table of $x$,$y$-coordinates in the unit of pixel, which then can be converted to other measurement, such as $\mu$m, according to the resolution of the camera. The *sojourner* package provides methods to import/export, process, and analyze such data. Analysis techniques include the diffusion coefficient using the mean square displacement (MSD), displacement cumulative distribution function (CDF), as well as hidden Markov model (HMM) methods (in next release) from such input data. This vignette covers basic usage of the package. ## Installation Using BiocManager: ```{r, eval = FALSE} if(!requireNamespace("BiocManager", quietly = TRUE)) install.packages("BiocManager") BiocManager::install("sojourner") ``` Or, using the release version from GitHub: ```{r, eval = FALSE} devtools::install_github("sheng-liu/sojourner", build_vignettes = TRUE) ``` To load the package: ```{r, eval=TRUE, messsage = FALSE, warning=FALSE, results = 'hide'} library(sojourner) ``` # Basic Track Data Structure The input file for *sojourner* is the Diatrack (.txt or .mat), ImageJ Particle Tracker (.csv), SLIMfast (.txt), or u-track (.mat) output file. Tracks are extracted from the output files and then stored in a `trackll`, a list of track lists. A track denotes a single trajectory stored in a `data.frame` recording $x$, $y$, $z$ and (video) frame number. A `trackl` track list is a collection of track data from one video file output. Lastly, a "trackll" is a collection of "trackl" in one folder. In this example, the folder "SWR1_2" contains one track list with 207 tracks and one track list with 139 tracks: ```{r, eval = TRUE, results = 'hide'} folder=system.file("extdata","SWR1_2",package="sojourner") trackll = createTrackll(folder=folder, input=3) ``` ```{r, eval = TRUE} str(trackll,max.level=2, list.len=2) ``` Multiple folders can also be stored in a single trackll. Here, `trackll` denotes a list of folders, and each `trackl` contains all tracks from every file in the folder. This structure can be constructed as such: ```{r, eval = FALSE} trackll=list(FOLDER=list(track=track)) ``` ```{r, eval = FALSE} # Construct trackll from data.frame trackl=list(track_1=dataframe_1, track_2=dataframe_2, track_3=dataframe_3, ... track_n=dataframe_n) trackll=list(FOLDER_1=trackl_1, FOLDER_2=trackl_2, FOLDER_3=trackl_3, ... FOLDER_n=trackl_n,) ``` Merging all tracks from every file in the folder into a single `trackl` in a `trackll` can also be done through `mergeTracks()`. In this example, both track lists from "SWR1_2" were merged together to form a list of 346 tracks. ```{r, eval = TRUE, results = 'hide'} folder=system.file("extdata","SWR1_2",package="sojourner") trackll = createTrackll(folder=folder, input=3) trackll <- mergeTracks(folder=folder, trackll=trackll) ``` ```{r, eval = TRUE} str(trackll,max.level=2, list.len=2) ``` To see the coordinates of an individual track: ```{r, eval = TRUE, warning=FALSE, echo = TRUE, results = 'hold'} # Specify the folder name and the track name trackll[["SWR1_2"]]["mage6.1.4.1.1"] # Alternatively, specify the index of the folder and the track # trackll[[1]][1] ``` # Track Data Manipulation *sojourner* provides tools to manage and manipulate track data. Below are a few examples of the commonly used functions to import, process, and export single molecule track data. ## Reading Track Data First, read in track data taken from Diatrack, ImageJ Particle Tracker, SLIMfast, or u-track. In this example, we read data from ImageJ .csv-formatted files from "SWR1_2". ```{r, eval=TRUE, warning=FALSE, echo=TRUE, results = 'hold'} # Designate a folder and then create trackll from ImageJ .csv data folder=system.file("extdata","SWR1_2",package="sojourner") trackll = createTrackll(folder=folder, input=3) # Alternatively, use interact to open file browser and select input data type # trackll = createTrackll(interact = TRUE) ``` ## Processing Track Data ### Linking Skipped Frames Different tracking softwares (Diatrack, ImageJ Particle Tracker, SLIMfast, u-track) often use varying algorithms to stitch together single molecule localizations into trajectories. Sometimes, a molecule can disappear for several frames and can classify single trajectories as two. `linkSkippedFrames()` allows users to link trajectories that skip (or do not appear for) a number of frames. In this example, all tracks that skip a maximum of 10 frames and reappear within 5 pixels are linked: ```{r, eval=TRUE, warning=FALSE, echo = TRUE} # Basic function call of linkSkippedFrames trackll.linked <- linkSkippedFrames(trackll, tolerance = 5, maxSkip = 10) ``` ### Filtering Tracks by Length `filterTrack()` can be used to filter out track that have lengths (defined by number of frames) that fall within a specified range. In this example, only tracks that appear for at least 7 frames are kept: ```{r, eval=TRUE, warning=FALSE, echo = TRUE} trackll.filter=filterTrack(trackll ,filter=c(7,Inf)) # See the min and max length of the trackll # trackLength() is a helper function output track length of trackll lapply(trackLength(trackll),min) lapply(trackLength(trackll.filter),min) ``` ### Trim Tracks to a Length `trimTrack()` is used to to trim/cutoff tracks. Given a specified range of track lengths (defined by number of frames), only tracks that fall within the range are kept, otherwise trimmed to the upper limit. In this example, all tracks with lengths greater than 20 have their lengths trimmed to 20: ```{r, eval=TRUE, warning=FALSE, echo = TRUE} trackll.trim=trimTrack(trackll,trimmer=c(1,20)) # See the min and max length of the trackll # trackLength() is a helper function output track length of trackll lapply(trackLength(trackll),max) lapply(trackLength(trackll.trim),max) ``` ### Applying Image Masks In order to apply a binary image mask to track data, one can use `maskTracks()`. In this example, we use image masks for each track file generated by thresholding the initial fluorescent glow of the nucleus. See man pages for `indexCell()`, `filterOnCell()`, and `sampleTracks()` for more advanced functionality (e.g. separating tracks, filtering, and sampling by masked cells). ```{r, eval=TRUE, warning=FALSE, echo=TRUE, results = 'hide'} # Basic masking with folder path with image masks folder = system.file("extdata","ImageJ",package="sojourner") trackll = createTrackll(folder, input = 3) trackll.filter=filterTrack(trackll ,filter=c(7,Inf)) trackll.masked <- maskTracks(folder = folder, trackll = trackll.filter) ``` To plot the masks: ```{r, eval=TRUE, warning=FALSE, echo=TRUE, results = 'hide',fig.show="hide"} # Plot mask plotMask(folder) ``` To plot the nuclear overlay of the un-masked data: ```{r, eval=TRUE, warning=FALSE, echo=TRUE, results = 'hide',fig.show="hide"} # If nuclear image is available plotNucTrackOverlay(folder=folder,trackll=trackll) ``` To plot the nuclear overlay of the masked data: ```{r, eval=TRUE, warning=FALSE, echo=TRUE, results = 'hide',fig.show="hide"} # If nuclear image is available plotNucTrackOverlay(folder=folder,trackll=trackll.masked) ``` ## Exporting Track Data To export a `trackll`, use `exportTrackll()` to save track data into a .csv file in the working directory. This function saves the data into the same format used by ImageJ Particle Tracker, fully preserving all track information, and maintaining short read/write computation time and readability in Excel/etc. ```{r, eval=FALSE} # Basic function call to exportTrackll into working directory exportTrackll(trackll) # Read exported trackll saved in working directory trackll.2 <- createTrackll(folder = getwd(), input = 3) ``` # Plotting Tracks The first thing one may want to do is to see how the tracks look like in 2-D space. The track name is useful in this case if one wants to see specific trajectories and its associated movies. All one needs to do is to create a .csv file contains trajectory names in its first column. *sojourner* package contains such an example .csv file. ```{r, eval = FALSE} # specify the path of the file containing trajectory index names, index file index.file2=system.file("extdata","INDEX","indexFile2.csv", package="sojourner") # specify the folders containing the output files folder1=system.file("extdata","SWR1",package="sojourner") folder2=system.file("extdata","HTZ1",package="sojourner") # plot trajectories specified in the trajectory index file plotTrackFromIndex(index.file=index.file2, movie.folder=c(folder1,folder2),input=3) ``` The output plots the tracks based on its name, with the information contained in its name (i.e. start frame and length/duration), one can also pull out its movie. See `? plotTrack` for more plotting options. # Distribution of Trajectory Lengths One may be interested to see the distribution of the (time) length of the tracks. This can be done by calculating and plotting the dwell times: ```{r, eval = TRUE, warning=FALSE} dwellTime(trackll,plot=TRUE) # default t.interval=10, x.scale=c(0,250) ``` # Calculating Diffusion Coefficient Using MSD-Based Method Below is an example of simple analysis workflow using diffusion coefficient. ## Data Import We will need to first create a basic data structure (trackll, i.e. list of track list) for all analysis. It contains the trajectory information of all tracks in a specified folder, using createTrackll() function. ```{r, eval = TRUE, warning=FALSE, echo=TRUE, results = 'hide'} # Specify folder with data folder=system.file("extdata","HSF",package="sojourner") # Create track list trackll<-createTrackll(folder=folder, interact=FALSE, input=3, ab.track=FALSE, cores=1, frameRecord=TRUE) # Take a look at the list by str(trackll,1) ``` ## Data Clean Up We then clean up the data with filters and masks. In this case, we select only the tracks with 3 frames or more, and remove tracks that is outside of the nuclei. ```{r, eval = TRUE, warning=FALSE, echo=TRUE, results = 'hide'} # Filter/choose tracks 3 frames or longer for all analysis trackll.fi<-filterTrack(trackll=trackll, filter=c(min=3,max=Inf)) # Apply mask and remove tracks outside nuclei trackll.fi.ma<-maskTracks(folder,trackll.fi) ``` ## Visulization of Trajectories To see how the tracks look like, we can plot the tracks over cell image (e.g. bright field image or nuclei overall fluorescence) to see the distribution of tracks. ```{r, eval = TRUE, warning=FALSE, echo=TRUE, results = 'hide',fig.show="hide"} # Overlay all tracks on nuclei plotNucTrackOverlay(folder=folder,trackll=trackll.fi,cores=1, max.pixel=128,nrow=2,ncol=2,width=16,height=16) # Overlay tracks after nuclear mask plotNucTrackOverlay(folder=folder,trackll=trackll.fi.ma,cores=1, max.pixel=128,nrow=2,ncol=2,width=16,height=16) ``` ```{r, eval = TRUE, warning=FALSE, echo=TRUE, results = 'hide'} # Overlay tracks color coded for diffusion coefficient plotTrackOverlay_Dcoef(trackll=trackll.fi.ma, Dcoef.range=c(-2,1), rsquare=0.8, t.interval=0.01, dt=6, resolution=0.107) ``` We used Dcoef.range to select tracks whose Dcoef is within the range. Other tracks will not be plotted. ## Analysis We then move on to the analysis of the biophysical properties of a molecule, e.g. mean square displacement (MSD), diffusion coefficient (Dcoef), residence time (RT) etc. For an overall view of one factor, we may want to merge all tracks in a folder. ```{r, eval = TRUE, warning=FALSE, echo=TRUE, results = 'hide'} # Merge tracks from different image files in the folder trackll.fi.ma.me=c(mergeTracks(folder, trackll.fi.ma)) ``` And sometimes, we may have data of same type in multiple folders and want to combine them together. ```{r, eval = TRUE, warning=FALSE, echo=TRUE, results = 'hide'} folder2=system.file('extdata','HSF_2',package='sojourner') trackll2=createTrackll(folder2,input=2, cores = 1) trackll2=maskTracks(folder2,trackll2) trackll2=mergeTracks(folder2,trackll2) # Combine the tracklls together Trackll.combine=combineTrackll(trackll=c(trackll,trackll2),merged=TRUE) ``` Now we can calculate msd, Dcoef, and residence time of a molecule. ```{r, eval = TRUE, warning=FALSE, echo=TRUE, results = 'hide'} # calculate MSD for all tracks longer than 3 frames msd(Trackll.combine,dt=20,resolution=0.107,summarize=TRUE,cores=1,plot=TRUE,output=TRUE) ``` For some analysis, we may just want to a portion of the trajectory, this can be done by using trimTrack() function. ```{r, eval = TRUE, warning=FALSE, echo=TRUE, results = 'hide',fig.show="hide"} trackll.combine.trim=trimTrack(Trackll.combine,trimmer=c(1,11)) msd(trackll.combine.trim,dt=10,resolution=0.107,summarize=TRUE,cores=1,plot=TRUE,output=TRUE) ``` ```{r, eval = TRUE, warning=FALSE, echo=TRUE, results = 'hide'} Dcoef(trackll=trackll.combine.trim,dt=5, filter=c(min=6,max=Inf), method="static", plot=TRUE, output=TRUE) ``` ```{r, eval = TRUE, warning=FALSE, echo=TRUE, results = 'hide',fig.show="hide"} ## calculating displacement CDF # Calculate all dislacement frequency and displacement CDF for all tracks longer than 3 frames displacementCDF(trackll.combine.trim, dt=1, plot=TRUE, output=TRUE) ``` ```{r, eval = TRUE, warning=FALSE, echo=TRUE, results = 'hide',fig.show="hide"} ## Analysis for Residence time. # Calculate 1-CDF (survival probability) of the dwell time/residence time of all tracks (not trimmed) longer than 3 frames compare_RT_CDF(trackll= Trackll.combine, x.max=100, filter=c(min=3,max=Inf), t.interval=0.5, output=FALSE) ``` ```{r, eval = TRUE, warning=FALSE, echo=TRUE, results = 'hide'} # Calculate residence time of tracks by 2-component exponential decay fitting of the 1-CDF curve fitRT(trackll= trackll.fi.ma.me, x.max=100, N.min=1.5, t.interval=0.5) ``` This is just a brief example of analysis one can do using the statistic tools provided in sojourner package. Further information regarding function usage or analysis techniques can be found in function help docs and on https://sheng-liu.github.io/sojourner/ website. # sojournerGUI: A Shiny Interface A Shiny app implementation of many of the core features of sojourner. Namely, the basic abilities of reading trackll video files (of all supported types), processing tracks (linking, filtering, trimming, masking, merging), and analyzing tracks (MSD, Dcoef, CDF, and dwell time). The application interface provides a code-free GUI, suited with dynamic and interactive plots, that is relatively easy to use. The app is still in alpha development and only supports the base functions needed for educational capabilities and such. A command history log, named *command_history.R* in the working directory, will be continuously updated each time a command is called for diagnostic, replication, and tracking purposes. ## Launching ```{r, eval=FALSE, echo=TRUE, results='hide', warning=FALSE} sojournerGUI() ``` ## Helpful Tips * If in doubt, check the console output of each command inputted through the GUI. These will show error and warning messages as needed. * Normal distribution, compare folder, and kernel density masking features are currently not supported. * Reading tracks depends on a **running session of R**, as it uses its native `file.choose()` function. # SessionInfo Here is the sessionInfo() on which this document was compiled: ```{r, eval = TRUE} sessionInfo() ```