Seismic-rna

- When version 0.1.2 enabled dropping duplicate read names, it inadvertently introduced a bug that caused subclasses of `BitAccum` to crash when checking for duplicate read names. This problem has been fixed.
- More detailed information on installation has been added to `INSTALL.md` and `VENV.md`.

**Full Changelog**: https://github.com/rouskinlab/seismic-rna/compare/v0.1.2...v0.1.3

- The mask, cluster, and table steps now drop any read whose name matches the name of a previously-seen read. The first read given for each unique name is kept, and a warning message including the names of the dropped reads is issued. In earlier versions of the software, duplicate read names instead caused an error to be raised.

- The default mode of alignment has changed from end-to-end (which is still the default in Bowtie2 v2.5.1) to local. The lab decided on this change upon realizing that end-to-end mode can introduce large numbers of spurious mutations at the ends of reads prepared using random fragmentation and/or random priming, even following adapter trimming. Hence, we recommend end-to-end mode for and only for samples prepared as PCR amplicons with sequence-specific primers (for which local mode misses mutations within several bases of the ends of the amplicons). The end-to-end mode can still be enabled using the flag `--bt2-end-to-end`.
- The unit tests in the `core` subpackage have been split into one module of tests per module of code. Previously, they were all in the module `core.test`, which was becoming too large to develop. Unit tests of the other subpackages will likely be reorganized in the same manner in future versions.

Features

This release introduces the essential utilities of SEISMIC-RNA:

- Alignment
- Relation
- Masking
- Clustering
- Tabulation
- Folding*
- Graphing*

*These utilities are implemented for related and masked data, but not yet for clustered data, on which they will fail or produce nonsensical output. This issue will be fixed in a future release.


Caveats

- This target audience of this release is members and collaborators of the Rouskin Lab (i.e. people with whom I already work, who can ask me for all the tech support they wish).
- I have not started writing the documentation meant for end users. So far, I have merely annotated the source code with docstrings and comments. I will add the documentation in future releases.
- Graphing features are very limited so far. The only supported graphs are counts and fractions of mutations per position in a reference sequence, and only from the relate and mask steps. More types of graphs and support for clustering will be added in a future release.
- The relate step is moderately slow. It will be rewritten in C in a future release.
- The table step is very slow. It will be optimized if feasible in a future release.
- The unit tests so far cover only a small fraction of the entire code, with a focus on the relate step, which has the most complicated pure-Python algorithm in the project. That said, my own end-to-end tests have revealed no significant bugs in the current release.
- I do not consider this release to be sufficiently complete, documented, or tested to share with researchers outside of the Rouskin Lab.