Mikado

Enhancement release. Following version 1.2.3, now Mikado can accept BED12 files as input for convert, compare and stats (see [122](https://github.com/EI-CoreBioinformatics/mikado/issues/122)). This is becoming necessary as many long-reads alignment tools are preferentially outputting (or can be easily converted to) this format.

Mainly this is a bug fix release. It has a key advancement though, as now Mikado can accept BED12 files as input assemblies. This makes it compatible with Minimap2 PAF > BED12 system.

Minor bugfixes:
- Now Daijin should handle correctly the lack of DRMAA
- Now Dajin should treat correctly single-end short reads

Highlights for this version:

- The version of the algorithm for retained introns introduced in 1.1 was too stringent compared to previous versions. The code has been updated so that the new version of Mikado will produce results comparable to those of versions 1 and earlier. **ALL MIKADO USERS ARE ADVISED TO UPDATE THE SOFTWARE**.
- Daijin now supports Scallop.
- Now Mikado will print out also the alias in the final picking tables, to simplify lookup of final Mikado models with their original assembly (previously, the table for the .loci only contained the Mikado ID).
- Various changes on the BED12 internal representation. Now Mikado can also convert a genomic BED12 into a transcriptomic BED12.
- Updated the documentation, including a tutorial on how to create scoring files, and how to adapt Daijin to different user cases.
- Now finalised transcripts will always contain a dictionary containing the phases of the various CDS exons.
- Mikado prepare now will always reverse the strand for mixed-splicing events.
- Added unit-tests to keep in check the regression in calling retained introns, and for the new BED12 features.
- Minor bugfixes.

Highlights for this release are the swithing by default to Prodigal in lieu of TransDecoder and to DIAMOND instead of NCBI BLASTX. The rationale behind the change is that the two former default programs scale poorly with the size of datasets, as neither was designed to maintain a good execution speed with potentially million sequences. Prodigal and DIAMOND fare much better with big datasets, and do speed up significantly the execution of the whole Daijin pipeline.

Changes in this release:

- Mikado is now compatible with NetworkX v. 2x.
- Mikado now accepts ORFs calculated by Prodigal, in GFF3 format, instead of only those by TransDecoder in BED format.
- Mikado compare indices now are **SQLite3 databases**, not compressed JSON files as in previous versions. This should allows for a faster loading and potentially, down the line, the chance to parallelise compare.
- By default, Daijin now uses **Prodigal and DIAMOND** instead of TransDecoder and BLAST. This should lead to massive speed-ups during the pipeline, although at the cost of slightly reduced accuracy.
- Improved the algorithm for finding retained introns, using a graph structure instead of potentially checking against every other transcript in the locus.
- Mikado configure now has a new flag, "--daijin", which instructs the program to create a Daijin-compatible configuration file, rather than a Mikado-only one.
- Fixed some bugs in Daijin regarding the handling of Long Reads.
- Fixed a bug in Daijin regarding the calculation of Trinity parameters - previously, Daijin could potentially ask Trinity for parameters for N times, where N is the number of required assemblies, lengthening the startup time.
- Solved a bug that created incompatibility with BioPython >= 1.69
- Solved some bugs that prevented Daijin from functioning correctly on a local computer
- Now Daijin by default recovers the information to load external software from an external configuration file. This allows for having a standard configuration file for external programs, without having to modify the main configuration all the time.
- Now Daijin creates and/or load a standard configuration file for the cluster, "daijin_hpc.yaml".

BugFix release.

- Fixed a bug which caused Mikado to go out of memory with very dense loci, when calculating the AS events.
- Fixed a bug which caused the log not to be saved correctly during the indexing for Mikado compare.
- Fixed a bug which caused Mikado pick to crash at the end, on rare cases.
- Data to be transmitted to picking process children is now stored in a temporary SQLITE3 database, to lessen memory usage and queue hangups.
- Fixed a bug while that caused a crash while Mikado was trying to approximate complex loci.
- Switched away from using clique-based algorithms, as they tend to be very memory intensive.