Table 2. Marker sets resulting from three sequential iterations of the CT analysis method.

Marker identity	Marker count
CT iteration 1: 88.9% accuracy
PTX3, pentaxin 3	35
ILK, integrin linked kinase	34
ENTPD6, ectonucleoside triphosphate diphosphohydrolase 6	1
GPCR CG50207	1
CT iteration 2: 72.2% accuracy
SFRS7, splicing factor, arginine/serine-rich 7	34
ENTPD6, ectonucleoside triphosphate diphosphohydrolase 6	24
CBRC7TM_424 GPCR	1
APAF-1, apoptotic protease activating factor 1	1
ERMAP, erythroblast membrane-associated protein	8
CGFLC_31120	1
GPCR CG50207	1
LDHA, lactate dehydrogenase A	1
CT iteration 3: 80.6% accuracy
LYPLA1, lysophospholipase 1	34
GPCR CG50207	26
CBRC7TM_424, GPCR	1
APAF-1, apoptotic protease activating factor 1	8
CGFLC_31120	1
LDHA, lactate dehydrogenase A	1

For each iteration, the two most frequent markers from the previous iteration were deleted from the expression data training set. Percent accuracy, the proportion of drug treatments correctly predicted by the respective marker set. G protein-coupled receptor (GPCR) CG50207 and CGFLC_31120 are novel sequences of GPCR and unknown function, respectively. Marker count, number of times out of the 36 model building episodes a particular gene was selected as a marker