Figure 1
From: SPACE: an algorithm to predict and quantify alternatively spliced isoforms using microarrays
Influence of noise on the estimation of relative transcript concentrations of BIRC5 gene (synthetic data). BIRC5 gene structure is shown in Figure 6a and in Additional file 1 (Figure S4). (a) Additive noise effect on estimation of relative transcript concentrations. The y-axis shows the MAE between the relative concentration of transcripts without noise and that estimated by the algorithm under the effect of different degrees of additive noise (MAE %). Additive noise is in the form of y + ε with ε ~N (0, σ ε 2). The units of the x-axis are the variances σ ε 2 of the additive error added to the simulated concentrations (10, 100, 1,000, 10,000, 100,000). These variances represent roughly 0.5%, 2%, 5%, 15% and 50% of the energy of the signal, respectively. (b) Multiplicative noise effect on estimation of relative transcript concentrations. The y-axis shows the MAE between simulated and estimated relative concentrations under the effect of different degrees of multiplicative noise (MAE %). Multiplicative noise is in the form of y·eηwith η ~N (0, σ η 2). The units of the x-axis represent the variances σ η 2 of the multiplicative error (5 × 10-5, 0.0005, 0.005, 0.05, 0.5). These variances represent roughly 0.7%, 2%, 7%, 25% and 100% of the energy of the signal, respectively. The different degrees of additive and multiplicative noise are tested while the other parameters are in the 'central point' condition (40 arrays and probes at exons and junctions). This means that there is always a component of additive and multiplicative noise in the form of y·eη+ ε. Errors are represented by boxplots. A boxplot is a graphical representation of the variability of a random signal. They are composed by a box and a whisker. The box extends from the lower quartile to the upper quartile values and there is an additional horizontal line that shows the median. The whiskers are vertical lines extending from each end of the boxes to show the extent of the rest of the data. Outliers are data with values beyond the ends of the whiskers and are represented by crosses.