For <=v0.4.0, when non CG contexts were requested, both minimod view and freq were reporting modifications from erroneous contexts. We have fixed and tested this issue from >= v0.5.0.
For <=v0.4.0, we allowed primary, secondary, supplementary alignments when viewing and calculating frequencies. From >=v0.5.0, we only consider primary and supplementary alignments by default. We introduced –allow-secondary option to enable considering secondary alignments. Minimod still errors out if hard-clipping is found.
For <=v0.4.0, when –insertions option is used, erroneous reference positions were included in both freq and view outputs. We have fixed it in >=v0.5.0.
For <=v0.4.0, status of skipped bases (low probability C+m. or unknown C+m? in MM tag) were not handled. We have fixed it in >=v0.5.0. When skipped as low probability, minimod consider them to have 0 modification probability. Otherwise, when skipped as unknown, they are ignored.
For >=v0.5.0, mod_prob in the output of view subtool will slightly change due to a change on how probability conversion from unsigned 8-bit N to float probability p. Previously, it was p = N/255.0 and now we use p = (N+0.5)/256.0 as the conversion function.
For >=v0.5.0, when a context is provided (ex: -c h[CG]) option, we output modifications where the modified read base matches aligned reference base. However, when the context is * (ex: -c a[*]), this comparison is ignored and modifications at both matched and mismatched positions are output. Note that only the modified base is compared with reference base, not the whole context.
For >=v0.5.0, we print a warning if a certain modification and a context is untested. For versions before this, only m[CG] for DNA (genome) and a[A] for transcriptome were tested.
For <=v0.4.0, minimod view output was not properly sorted when using multiple threads. We have fixed it >=v0.5.0 and now view output is ordered in the same as the order the reads appear in the input BAM file, and for each read, entries are sorted by reference contig, reference position, strand, and modification code.