v1.2.0

Major speed improvement via a complete re-implementation (from start to finish) in C++.
Some arguments have been re-named, please see README for further details.

v1.1.3

Release v1.1.3 of transigner includes significant changes.

Updates

multi-threading and partial C implementation

We implemented the EM module in C and enabled multi-threading using OpenMP. Linux systems with OpenMP will now be able to take advantage of much faster execution time.

position-specific weights mode

By default, transigner uses alignment scores to guide the EM iterations. In the new position-specific weights mode, we compute the likelihood that a read covers the aligned bases that gets combined with alignment scores for guiding the E-step updates.

sampling-based push algorithm

Previously, transigner converted soft, fractional read assignments to hard, 1-to-1 assignments by assigning reads to transcripts with the most read fractions allocated to. We've revised the algorithm to instead sample from weighted discrete distributions with probabilities equivalent to the read fractions associated with each transcripts (i.e., random variables in this context).

minor changes

• Alignment end distance-based filters are now optional.
• easy.sh script was added to support seamless execution of transigner's modules.

v1.1.1

Minor bug fix for the EM module. Malfunctioning of the --drop flag.

Release v1.1.0

Major restructuring for all modules. Remove jf_aligner relevant codes. Also released on PyPI to enable pip installation.

Release v1.0.1

Initial release of TranSigner