Expression profile of protein coding genes (PCGs) in unilineage culture of cord blood HPCs. (A) Gene microarrays were representative of a 54630 redundant set of sequences having linear normalized signal values ranging from 0 to 71337. Only sequences having at least one value >500 (an under-expression arbitrary threshold roughly corresponding to 1% of maximum value) were considered significantly expressed (Expressed Sequences) and sub-grouped according to two different criteria: i) expression level, by using a second arbitrary threshold (5000) in signal values that results in Medium Expressed (maximum value in the range 500-5000) and High Expressed (at least one value >5000) genes; ii) fold increase/decrease calculated as the ratio of differentiated cells (tmax) to CD34+HPCs (t0) signal, obtaining Changed, Medium Changed and High Changed subsets of sequences showing in at least a lineage a signal fold change >3×, 3-6× or >6× respectively. (B) Self-organizing maps of global gene expression for the E, G, MK and Mo differentiation pathways. The array data set of N = 17655 Expressed Sequences, multiple randomly sampled 1000 or 100 sequence subsets and same-sized Random Numbers sets were clustered using the GEDI software into nearly-squared grids composed by miniclusters. On the left data or number set analysed (progressive number) for each line is defined, on the top of each column lineage and culture time are indicated. Below each grid the Pearson correlation with CD34+ (or with the first sample for Random Numbers) is reported. In the grids each pixel corresponds to a minicluster that is located at the same position in all the grids of the same line and is composed by a variable small group of sequences sharing a similar expression pattern, whose number is indicated by the Gene Density Map. On the right grid size (pixel number) is indicated. Colour of each minicluster indicates the mean of log2-transformed expression value of corresponding sequences or Random Numbers, as indicated on the column bar on the right.