ProactiveCoreDefense

Protect your flow cytometry core from bad data.

Autonomously monitor user sample quality for flow cytometers.

This is an "improved" version of the 'Sample Quality Monitor' repository I developed at Sanger (https://github.com/hally166/SampleQualityMonitor).
The primary differences are:

• Moved most of the workload into functions to help with portability across instruments
• Removed the Sanger specific statistics
• Moved most of the variables to the first lines to help with setup
• Removed the ability to email the core. This was very problematic and insecure. Now the data is saved to an Excel file and the image output is the default one from flowCut

The idea here is to test user sample quality by checking time vs fluorescence, as we do during analysis, but at the time of acquisition. This allows the core facility to be proactive helping our users spot problematic experiments and allows us to check for machine issues, such as recurrent blockages.

It finds the fcs files produced since that last operation and runs them through the R package flowCut (https://github.com/jmeskas/flowCut) which looks for deviations in fluorescence over time. The script then plots and records the “bad” files and saves the data to a spreadsheet.

The output currently looks like this.

Instructions

You run the script on the flow cytometer PC and save the data to the local network drive or to the PC. I prefer the network drive and I make sure that all my PCs map the drives with the same drive letter.

On the flow cytometer PC.

• Install R
• Open R from the start menu and install the required R packages:

install.packages("devtools")
devtools::install_github("jmeskas/flowCut")
install.packages("openxlsx")

You may be asked for some user input, I normally select 'no'.

On the network or flow cytometer PC

• Download the QC Script R file and ps1 file and save them to the network or local PC
• Open the file and change the options at the top. It is explained in the script
• Now the hardest part. Flow Cytometer manufacturers are incapable of producing consistent fcs files. One of the options is to record and to count the number of users, which is very useful. Sadly, the manufacturers record this is different ways {sigh}. Look for the following section and you will see some parts commented out using the # character. Add or remove these depending on your instrument.

#function to perform the flowCut procedure and return the results
flowCut_data<- function(ff){

By default I have left the $OP keyword in, but if you have a BD machine you will need to add a # to the start of this line and remove the one on the line that uses the 'EXPORT USER NAME' keyword. If you are having issues here, or if the number of images and rows on the spreadsheet don't match, then you need to load an FCS file and work out which keyword is being used.

• Save the file and close it.
• Open the ps1 file (PowerShell) and change the R path to where yours is and the R script path to where that is too. Save the file and close.
• Go to task scheduler in Windows and add a new daily task that runs the R script though PowerShell.exe with this arguemnt (or similar)

-ExecutionPolicy ByPass -File Q:\User_QC\QCScriptPwrShell.ps1

Testing

There are two ways to test that it works; using RStudio or running it though the command line. To run it in the command line click on 'Start' type 'CMD' and open the command line. Then navigate to the PowerScript file and run it

PowerShell.exe -ExecutionPolicy ByPass -File Q:\User_QC\QCScriptPwrShell.ps1

If this does not work, try running it in RStudio and do some troubleshooting. I'll make a FAQ and video later.

You should end up with an excel spreadsheet with the data including links to the image files and a new folder with the images inside.

You can deploy this multiple times from the same network location, but you will need to make multiple copies of the R and ps1 file so that you have one set per machine. Originally I had this work automatically by taking the $CYT keyword, but there are inconstancies across manufacturers in the use of FCS file keywords and it would have caused issues if you had not set up unique names for each of your instruments.

TODO

• Probably make it so only problamatic files are listed in the spreadsheet.
• The email feature was pretty cool, I'll work on that.
• Installer
• Real-time