Open skurscheid opened 7 years ago
(just using this to keep track of thoughts/ideas for the project, so that it is all in one place)
Contrary to what I had thought, ChIPsim allows to include an error model when simulating reads. Specifically, per-nucleotide read qualities can be simulated by uniform-randomly sampling Phred quality scores. Then ChIPsim uses an error model to perform nucleotide substitutions based on the simulated quality and a set of substitution probabilities. This gives leads to a set of reads with errors plus qualities.
We could essentially use a perl/python/awk/R script to sample Illumina base call quality strings from existing FASTQ data to build up an empirical distribution of read qualities and use this as the pool to sample random quality scores from...