We hypothesized that this was caused by sequencing error introduced into the simulated reads.
We mapped the simulated reads against the coding domain sequences from which they were derived and calculated mapping error rates.
While all reads mapped, the error rate was higher for reads that were predicted to be non-coding than those predicted to be coding (Welch Two Sample t-test, estimate = 0.00523, p < 0.001).
olga commented: Is there a figure for read error rates in coding vs noncoding?
olga commented: Is there a figure for read error rates in coding vs noncoding?