Current limitations

Known limitations and forecasts for when they will be removed.

1. During modkit pileup, it is assumed that each read should only have one primary alignment. If a read has more than one primary alignment it will be double counted, it is therefore recommended to remove duplicates before running Modkit.
2. The MAP-based p-value metric (details) performs a test that there is a difference in modification (of any kind) between two conditions. If a position has multiple base modification calls (such as 5hmC and 5mC) the calls are summed together into a single “modified” count. If a position differs only in the modification type (such as one condition has more 5hmC and the other has more 5mC) this effect will not be captured in the MAP-based p-value. The likelihood ratio test does capture changes in modification type.
3. The MAP-based p-value is not available when performing DMR on regions. This is because there is potentially large variability in the number of modified bases and coverage over regions. This variability translates into varying degrees of statistical power and makes comparisons difficult to interpret.