JASA Flow Cytometry Data


The flow cytometry data used in the JASA paper were derived from the experimental data posted by Sachs et al as a supplement to their Science paper. For our experiments there was some minimal processing done to discretize the data as well as prepare the database for use with the ‘bnslx’ package.

Nine of the total 12 datasets were used with 600 samples taken randomly from each dataset:

  1. 1.CD3 CD28

  2. 2.CD3 CD28 + iCAM2

  3. 3.CD3 CD28 + AktInhib

  4. 4.CD3 CD28 + G0076

  5. 5.CD3 CD28 + Psitect

  6. 6.CD3 CD28 + U0126

  7. 7.CD3 CD28 + Ly

  8. 8.PMA

  9. 9.B2cAMP

Each of these was log-transformed and then discretized into three levels using simple 1-d k-Means clustering. The dataset and the experiment identifiers are available here in both R and text formats. The BDE files contain the processed data while the all.Rda contains an R image of the original unprocessed data.

The graphs were constructed using the bnslx package (here) and the graphs constructed using databases built with the fill.db function (here) and then processed with the equi-energy sampler (here). A parallel tempering sampler is also available as the function “fast.order.pt,” which is often useful for tuning of the Equi-Energy sampler.