Hi,

One of the major novelty in disco 2.2.0 is that it automatically detects the threshold coverage under which k-mers are removed as they are considered as containing a sequencing error.

In 2.1.7 this value is 3 by default. In 2.2.0, this value can be found in the log (line starting by "thresholds").

Regarding your results, two possibility (co-)exist:

• variants predicted only by 2.1.7 are in fact generated by sequencing errors, they are false positives and it's a good thing they were non predicted in 2.2.0
• the automatic threshold detection was too stringent and, by consequence, real k-mers were removed and associated variants were not found.

This is difficult to conclude without a deeper look into your study details, the automatic c value, and the data complexity and coverage.

Pierre