Supplementary note 1. Identification of transcripts with dose-dependent expression change The transcripts whose expression changes are in accordance with the imatinib doses across sublines can fall into two categories, directly or inversely correlated with escalatinb imatinib doses. We used Bartholomew’s homogeneity law for ordered alternatives, modified F statistics for comparison of more than two groups.1 This test is an extended version of one-side t-tests modified to deal with more than two groups. First, 5 mean expression levels of control K562 and 4 resistant sublines were considered as continuum corresponding 200 nM of imatinib dose increments up to 800 nM. The significance level of up- and down-progression was determined under the alternative hypotheses, H1 : mcont m200nM m400nM m600nM m800nM or the vice versa, H1 : mcont m200nM m400nM m600nM m800nM , respectively. Five mean values of expression ( mcont , m200nM , m400nM , m600nM , and m800nM ) of a gene are reduced into a set of l means ( m1 , m2 , m3 , m4 , m5 ) treating m j , j 1 (a j m j a j 1m j 1 ) (a j a j 1 ) as a single observation when m j m j 1 or m j m j 1 considering weights a j and a j 1 . Then, Fk is calculated using reduced l number of means: ni (mˆ i x..) 2 Fk (l 1) S 'e2 i 1 k where, S 'e 2 is residual sum of squares with degree of freedom, N l . 1 The significance level is calculated using incomplete B function, k k l 2 l 2 Pr{Fk r} p(l , k ) Pr{Fl 1, N l r} p(l , k ) I1 z ( 12 ( N l ), 12 (l 1)) Pr{Fk 0} p(1, k ) where, z (l 1)r ( N l (l 1)r ) . Since we dealt with many number of probe sets, the adjustment problem arising from multiple hypotheses testing must be considered. We used Q-value package (http://faculty.washington.edu/~jstorey/qvalue/) that performs multiple adjusting based on false discovery rate (FDR) methods.2 FDR is measured as the ratio of truly null features over the total number of tests called significant and 5% of FDR (q-value < 0.05) was considered significant. The transcripts that passed the test for up- and downprogression were further tested for t-test both on high imatinib dose groups (600 nM and 800 nM) as previously described.3 We compared the gene expression between K562 control and high imatinib dose groups (600 nM and 800 nM) using Student’s t-test with Welch correction. Test was performed under the assumption of unequal variance, as previously suggested.4 Genes that passed the second criteria, both significant in 600 nM and 800 nM t-tests, were regarded to be coordinately changed across resistant sublines. In this way, genes whose expression changes correlated or inversely correlated with imatinib doses increments were identified and defined as up-progression and downprogression, respectively. 2 References 1. Bartholomew,D.J. (1959) A test of homogeneity for ordered alternatives. Biometrika, 46, 36-48. 2. Storey,J.D. and Tibshirani,R. (2003) Statistical significance for genomewide studies. Proc. Natl. Acad. Sci. U. S. A, 100, 9440-9445. 3. Tseng,Y.H., Butte,A.J., Kokkotou,E., Yechoor,V.K., Taniguchi,C.M., Kriauciunas,K.M., Cypess,A.M., Niinobe,M., Yoshikawa,K., Patti,M.E. et al. (2005) Prediction of preadipocyte differentiation by gene expression reveals role of insulin receptor substrates and necdin. Nat Cell Biol., 7, 601-611. 4.Ideker,T., Thorsson,V., Siegel,A.F. and Hood,L.E. (2000) Testing for differentiallyexpressed genes by maximum-likelihood analysis of microarray data. J. Comput. Biol., 7, 805-817. 3