Calico clusters ddPCR data automatically by boosting signals of positive droplet partitions. No hand drawing of boundaries!
Calico requires R to be installed. Get it from https://cran.r-project.org/
Download calico by clicking on the green "download/clone" button, then click "Download ZIP". Unzip the archive to the location of your choice. The calico.R script can then be copied to any location.
Calico is designed to be run as a standalone script. It does however require some third-party libraries from CRAN. To install them, just run calico by typing in the command prompt:
calico.R
or
calico.R -h
or
Rscript calico.R
In Windows you may need to specify the path of Rscript. For example:
"c:\Program Files\R\R-3.4.1\bin\Rscript.exe" calico.R
Alternatively in Windows, you can temporarily set the PATH environment variable to automatically point to where R is installed. Start the command line, and at the beginning of a command line session:
SET PATH=%PATH%;"C:\Program Files\R\R-3.4.1\bin\" or wherever R is installed.
A demo can be run by typing ./example.sh from the command line (Linux/Mac OS X) or double-clicking example_windows.bat. It will analyze an Evagreen-based multiplexed ddPCR dataset.
(You may need to directly call Rscript directly as described above. See "Installation" for an example.)
-
Create an empirical model using a positive control:
./calico.R -m -i example/data/20160923_MP_BRAFV600E_DNA_StdCurve_D03_Amplitude.csv -o example/results/model_20160923_MP_BRAFV600E_DNA_StdCurve -
Cluster an unknown dataset using the empirical model:
./calico.R -c -r example/results/model_20160923_MP_BRAFV600E_DNA_StdCurve -i example/data/20160923_MP_BRAFV600E_DNA_StdCurve_D04_Amplitude.csv -o example/results/20160923_MP_BRAFV600E_DNA_StdCurve_D04_Amplitude.clustered.csv >> example/results/log.txt -
(Optional) Draw a figure with the clustered dataset:
./calico.R -p -r example/results/model_20160923_MP_BRAFV600E_DNA_StdCurve -i example/results/20160923_MP_BRAFV600E_DNA_StdCurve_D04_Amplitude.clustered.csv -o example/results/20160923_MP_BRAFV600E_DNA_StdCurve_D04_Amplitude.clustered.png
The model generation step (-m) creates an output file that must be used for the clustering (-c) and plotting (-p) steps.
During the clustering (-c) stage, it creates an output csv file (each row is a droplet partition) with appended columns corresponding to the Mahalanobis score against each reference cluster, the final q-score, and the cluster call. It also outputs to stdout the filename, total number of droplets, and the number of droplets in each cluster. This line is useful when looping over many ddPCR assays, as it can be piped into a summary log file.
The plotting (-p) stage creates an output PNG with estimated boundaries from the reference model and associated cluster calls for the unknown dataset.
There are a number of parameters that you can set:
Quality score -q: this sets the quality score threshold for filtering out droplet partitions. Droplets with q-scores smaller than this value will be labelled as NA and omitted from analysis. This is used during the clustering (-c) stage.
Number of clusters -n: this sets the expected number of clusters from a postive control dataset. Default is 3, and has been tested for 4 with a Taqman ddPCR dataset. This is used during the model generation (-m) stage.
Tiles -t: this sets the tile size for the sliding window average. It creates a t x t sliding window to average over the gridded dataset. This is used during the model generation (-m) stage.