v3.0.3

This release addressed a few issues reported with the previous version, including:

• fix for cases where the --mix option was causing a segmentation fault with some data combinations
• preventing duplicate output transcripts for -e and --merge usage

v3.0.2 release

• Improved sensitivity and precision for long-read assembly use cases (-L and --mix).
  • Ida Shinder (@ishinder) contributed these improvements and fixes, including a feature that uses unaligned poly(A) tails to better anchor transcript ends.
• Addressed a rare regression where very complex splice graphs in long-read bundles with high error rates prevented StringTie from completing the assembly of those bundles.
• Addressed a long-standing issue where -e (expression estimation) did not always report all transcripts provided via -G when they did not overlap any reads (i.e., no coverage), which impacted downstream analyses using the prepDE script.
• Added a self-contained source package to the release (stringtie-*.offline.tar.gz), which includes test data and can be used to compile StringTie and run the included tests on systems without internet access.
• Added support for the Clang/LLVM C++ compiler toolchain on Linux.

v3.0.1 release

• fixed cases where -e option could not be used with the new -N/--nasc options
• increase floating point precision and numeric stability for calculation of coverage, FPKM and TPM values
• added a self-contained offline source package that can be built without online access
• added support for LLVM clang compilation

v3.0.0 release

• Significant improvements in assembly accuracy for both short and long RNAseq reads
• StringTie can now better recognize and assemble incompletely processed (nascent) transcripts which are abundant in Total RNA libraries (e.g. Ribo-Zero data)
• Addressed -e quantification issues where some transcripts with low coverage were no longer reported in the output (fixing #238, #357)

v2.2.3

Full Changelog: v2.2.2...v2.2.3

v2.2.2 patchfix release

• fix #348 : adds mutex to manage concurrent writing to covered referen… by @gsatas in #371
• fix #320 : fixes a out-of-bounds issue when a large number of predictions are generated
• fixes for other rare situations causing a program crash

Full Changelog: v2.2.1...v2.2.2

v2.2.1

• addressed an issue causing the -e option to not output low coverage transcripts
• fixed a --ptf data loading problem

v2.2.0

release 2.2.0

v2.1.7

• fix an occasional bug (crash) when --mix option was used

v2.1.6

• new --mix option added to use both short and long read alignments; when this option is used, the 2nd .bam file in the command line is expected to be a long reads alignment file (the 1st being the short-reads alignment file); -L option should not be used in this case.