Deregulation of erythroid marker expression in MDS erythroblasts. (A) Gene expression profiling. RNA from MDS with dyserythropoiesis (dysE; n = 5) versus normal (n = 4) cultured erythroblasts are hybridized on Affymetrix gene chip U133A. Neighborhood analysis of data from day 7 shows the probe sets predictive of MDS (red line) versus normal together with curves showing the 50% (black line), 5% (green line), and 1% (blue line) significance levels representing randomly permuted class distinctions. An under-expressed probe set (n = 265) predicts MDS (top panel) with a less than 5 in 100 chances of error, whereas no overexpressed probe set can make the distinction between MDS and normal (bottom panel). (B) RT-qPCR. Quantification of GYPA transcripts in the training and validation sets (9 MDS: ; 7 controls: □), pooled and expressed as mean normalized relative quantity (n_RQ) ± SEM (SE) to GAPDH. (C-D) Erythroid differentiation. (C) Representative dual fluorescence histograms of MDS with dysE (n = 30), control (n = 19), and MDS w/o dysE (n = 11) erythroblasts labeled with Abs to CD71 (FL1) and GPA (FL2). Percentages are indicated. (D) Ratio of median fluorescence intensity (RFI) of GPA to isotype per cell at day 14 expressed as box plots: lines within boxes represent the medians (Student t test for P values; *P < .05; **P < .01; ns indicates not significant).