TSAR Workflow by Shiny
## 1. Introduction TSAR, short for Thermal Shift Analysis in R, provides simple solution to qPCR data processing, computing thermal shift analysis given either raw fluorescent data or smoothed curves. The functions provide users with the protocol to conduct preliminary data checks and also expansive analysis on large scale of data. Furthermore, it showcases simple graphic presentation of analysis, generating clear box plot and line graphs given input of desired designs. Overall, TSAR Package offers a workflow easy to manage and visualize. TSAR Package is wrapped within a three separate shiny application regarding, data pre-processing, data analysis, and data visualization. All application can be opened in both interactive window or browsing engine by copy pasting server address into web browser.
2. Installation
3. Load TSAR Package
Use commands below to install TSAR package: library(BiocManager) BiocManager::install(“TSAR”)
library(TSAR)
library(shiny)
4. Data Pre-Processing
Load data and remove blank wells by specifying a range of wells with removerange = c() or individual wells with removelist = c(). User may also use function weed_raw() to open Shiny application to screen all raw curves and remove by curve selection. To propagate change to data to local change, close window properly by clicking Close Window. The updated data set will be stored in the global environment as new_raw_data.
data("qPCR_data1")
qPCR_data1 <- remove_raw(qPCR_data1, removerange = c("B", "H", "1", "12"))
runApp(weed_raw(qPCR_data1))
In the plot panel, user can interact with graph and select and un-select data by clicking on curves. User may also View Selected wells only and Remove Selected. To return to default page and current all change to data set, click Refresh Screening. To propagate the same changes to local data set, simple clickClose Window. Else, click Copy Well IDs or Copy Selected in Full Function Call to copy a remove_raw() call containing all selected wells into clipboard. Paste the call back into console or script and run to propagate changes locally. p.s. It is recommended to remove large areas of blank wells before calling weed_raw() to boost speed of application.
Selection by grid is also possible; click on grid to highlight wells of selection. Click again to un-highlight. Click too fast may lead to error due to speed of reaction, use Clear All Selected button at the top of page to reset selection. Note that clearing selections alters to-be-copied selections, but will not restore curves already removed in the graph. However, changes are not made permanently. If erroneously removed, simply close and reopen window to run again.
Use View Selected and Remove Selected to view and remove selection. To return to viewing rest of the data, click Refresh Screening. To propagate the same changes to local data set, click Copy Selected in Full Function Call to copy a remove_raw() call containing all selected wells into clipboard. Paste the call back into console or script and run to propagate changes locally. 
5. Data Analysis
Analyze data by calling function analyze_norm() and follow the workflow from top to bottom. Preview data table for changes occurring at each step and refer to graph to view fit of model on each well. Always refer to the message for hints and error references. A success message will also be prompted after each successful run of step.
Remember to save analysis output locally by clicking Save File. Always preview data before saving to ensure data contains all necessary information.
runApp(analyze_norm(qPCR_data1))
The top left panel output a preview of current data set. The right panel allows user to view the fit of model and Tm estimation by individual wells of selection. Once confirming correct data input and modeling effects,
Click Analyze all Wells to propagate model and analysis onto the rest of data. A preview of analyzed data will also be modeling and analyzing all 96 Wells will take between 30 to 50 seconds. If no modeling are needed, given data are smooth enough, analysis of all 96 wells should be completed under 5 seconds. p.s. actual time length many subject to change under different conditions
Upload well information by excel template and preview to confirm if information is correct. Use Preview to preview uploaded information and directly edit inside the window. User may also choose Manual Input to put in all condition information by Well. Make sure to hit Save Changes after editing and click Set Conditions to link all data to the conditions. A success message should be prompted. p.s. manually inputted information will override file upload. If mistakenly saved, please close window and re-run command.
Lastly, to save all analysis locally, click Save File after previewing output. p.s. make to select both under Choose dataset if intending to use graph_tsar() or other graphing tools within the package.
6. Data Visualization
Use function graph_tsar() to start a Shiny application for graphing options. Run na.omit() on data if error occurs. Four graphing options are allowed, boxplot of Tm, compare plots, and conditions plot.
Function takes optional data parameter. If analysis file is already imported in the environment, call function as graph_tsar(tsar_data). Refer to vignette “TSAR Workflow by Command” for instructions on how to merge using function merge_norm(). Else, user may use the merge data panel to upload and merge data of all test trials. Simply call graph_tsar() and click Upload and Merge Data button to reveal the panel to merge data.
Upload is limited to 30MB in size, but not count. After upload, user will be prompted corresponding numbers of input boxes to specify date of each experiment. Click Save Dates, then Merge and Save Data. A short preview of tsar_data will be prompted. For full lists and filters of well_ID and conditions_ID, refer to the helper buttons at the bottom of page.
Top panel outputs all plots, selected desired graphing features below and click generate to output graphs. p.s. Graphing compare plots and selected curves are takes longer than boxplot, please give it few seconds to load.
Boxplot has the option to be loaded interactive. Note interactive graphs have to be produced with legend combined.
