A question we occasionally get asked is why is the default good base improvement set to −20? The reasons are complex, but it is basically due to the way KB overcalls good bases. The purpose of the good base improvement setting is to allow processing of only those traces that are able to be improved by PeakTrace. This enables us to only charge you for the traces that are improved and not charge you for fail reactions, short PCR products, etc.
How can you determine which traces have been improved by PeakTrace? Since a KB predicted Q20+ base is not the same as a PeakTrace predicted Q20+ base, just counting predicted Q20+ bases does not allow you to know if the basecall is actually better or not (i.e. if the are more true Q20+ bases or not). There are three solutions to this problem:
- Add a correction factor to the PeakTrace quality score values (i.e. add 3 to every Q value in the Q17 to Q27 range). This way when you are comparing KB to PeakTrace predicted quality score counts you are making an apples-to-apples comparison. The downside of this approach is that PeakTrace would then overcall all the bases in the Q20 to Q30 range.
- Subtract a correction factor from the KB quality score values (i.e. minus 3 from every Q value in the Q20 to Q30 range). The downside of this approach is the reported Q20+ count would not match any external count, breaking downstream scripts and analysis.
- Set the comparison (good base improvement) so that it takes into account KB’s overcalling. This way a trace would be accurately identified as being improved by PeakTrace (i.e. have more real Q20+ bases), despite having fewer predicted Q20+ bases.
While none of these three approaches are ideal (the ideal would be for ABI to correct KB’s overcalling), the third approach is the least disruptive. Our analysis has identified −30 as the equivalence point for comparisons between KB and PeakTrace basecalled traces (i.e. a PeakTrace trace with ~30 fewer predicted Q20+ has as many real Q20+ bases as the original KB trace). Because we wish to be conservative, we have set the default for good base improvement to be −20 so that only traces that are improved by at least 10 real Q20+ bases are accepted. You can be even more conservative than the default value and set the good base improvement as high as −10 without getting charged for non-improved traces, however, any higher than this will result in you being charged.
The bottom line is that you shouldn’t set the good base improve higher than −10 if you want to get the best out PeakTrace.