Powered By Bing

Supplementary Information

Gene-expression patterns in drug resistant acute lymphoblastic leukemia cells and response to treatment.

Supplemental Table 4 : Number of significant probe sets discriminating drug resistance for each individual

Probe sets were rank-ordered according to their P-values, with the smallest P-values indicating the strongest statistical difference between resistant and sensitive patients. Among predefined significance levels (α), we chose the cut-offs that provided a low FDR rate (less than 10 percent when possible), high prediction accuracy and we limited the number of genes so that they did not exceed the number of samples, to facilitate future statistical analyses. The one exception was DNR (α =0.001, FDR=32%), because the FDR was never less than 10 percent. The final genes selected were required to be significant by both Wilcoxon rank sum test and t-test, using the selected P-value cut-offs (indicated in red in Supplemental Table 3). Listed below are the number of probe sets with a P-value lower than or equal to the value indicated (α), determined by Wilcoxon rank sum test and t-test, which yielded above for (a) all samples, including 153 probe sets (146 unique) and 107 known unique genes and 29 cDNA clones were selected by both methods for all four drugs; and (b) only the precursor-B lineage ALL, including 177 probe sets (172 unique) and 124 known unique genes and 28 cDNA clones were selected by both methods. Indicated in red are the numbers of top gene probes that were used to build the final drug resistance models.

a. All patients

 

Wicoxon rank
sum test

t-test

both tests

PRED

53*

49*

32*

VCR

76*

91*

54*

ASP

91+

81+

52+

DNR

27*

34*

15*

sum

   

153

b. B-lineage

Wicoxon rank
sum test

t-test

both tests

57+

69+

42+

74+

101+

59+

67

71

54

38*

50*

22*

   

177

α <0.0001, +α <0.0005, *α <0.001

< Previous | Next >