Smoothing read densities. (A,B) Local read heterogeneity of a single ERCC RNA in the 100% ERCC library (library 6). Smoothing read density using the GLM linear model (B), and the more complex MART model (C; see text). (D) Variance in read depth of randomly drawn 50-bp windows from all ERCC RNAs based on an unbiased simulation, raw data, and the smoothed coverage from the sequence specific models. (E) The effect of coverage on read depth variance in simulated data. For the most abundant quartile of transcripts (Q1, mean coverage >19.9), the ratio of the read depth of 50-bp windows to the average depth is between 0.2853 and 1.2360. For Q2, (mean coverage >1.5), inner quartile range for the ratio is between 0.1917 and 1.2540. For Q3, (mean coverage >0.09), it is 0 with moderate large outliers (>21). For Q4 (mean coverage <0.08), it is 0 with very large outliers (>400).
