Normalization bias analysis according to reference genes selection. (A) Example of normalized transcript levels of abf-B2 and xynC under different culture conditions (Thio-gentiobiose, Arbocel and C starvation samples) using 4 different Normalization Factors (NFs): NF(R2, R10, R6), NF(R5,R11,R3), NF(R3) and NF(β-tub). Log (base 2) of FC values on the Y axis (mean ± SD, n = 2 in this experiment) using glucose as the calibrator sample. (B) Comparison of NF(R2, R10, R6) to NF calculated from less stable genes, as well as from single genes such as R3 and β-tub. For each condition of interest (X axis, see Additional file 1), we calculated a normalization bias (i.e. under- or over-estimation of the normalised expression value of GOIs) as the ratio between the theoretically best NF (NF(R2, R10, R6) as determined from geNorm classification by using the entire set of conditions) and NF calculated from other combinations of reference genes. Log (base 2) of the normalization bias is represented on the Y axis. Yellow zone: less than 1.5 fold bias; Green zone: 1.5─2 fold bias; Blue zone: 2─3 fold bias; Red zone: 3─8 fold bias. (C) Quantile plot of the normalization bias values for each NF. The normalization bias (Log (base 2)) is represented on the X axis and the same colour code used in (B) was applied. The quantile fractions are represented on the Y axis. (♦) NF(R12,R7,R8), (☐) NF(R4,R9,R1), (◯) NF(R5,R11,R3), (+) NF(R3) and (×) NF(β-tub).