Each run was analysed individually to identify the accuracy of empirical quality scores. For each Ion Torrent quality score, we examined the error rate of bases assigned that score, and calculated the associated Phred score based on the error rate. To analyse the relationship between quality score and position within homopolymer, we sampled 20,000 reads from each run, and calculated the relative change in quality score from the first base to later bases within each homopolymer. We also recorded whether each of these consecutive bases was a correct call or an over-call.
