Rail-RNA v0.1.8a

Bug fix:

• Rare corrupted lines from first step that would break writing final alignments no longer appear.

Rail-RNA v0.1.8

Bug fix:

• Sorting no longer breaks for some versions/configurations of Coreutils sort. Previously, the tab separator used by Rail had not been specified explicitly in sort invocations. Thanks to Maximus B. for finding this bug.

Feature addition:

• elastic mode works with IAM roles, which are now required to launch EMR clusters. By default, these roles are read from the AWS CLI configuration file, but they may also be specified explicitly with the --instance-profile and --service-role command-line parameters.

Changes:

• The --no-consistent-view command-line parameter has become --consistent-view, and by default, EMR clusters are now launched with consistent view turned off. Consistent view is no longer needed in the US Standard zone because AWS recently introduced read-after-write consistency there.
• AMI 3.8.0 is now used by default in elastic mode.

Rail-RNA v0.1.7d

Bug fixes:

• A new compile of bedGraphToBigWig on UCSC's server had been incompatible with Rail-RNA. This version pegs the software to a particular version of bedGraphToBigWig hosted elsewhere.

Minor enhancements:

• Rail-RNA now accepts space-separated arguments of --indel-criteria, --intron-criteria, --deliverables/-d and --bowtie-idx/-x.
• Bowtie index basename arguments of --bowtie-idx/-x may now be specified in any order.

Rail-RNA v0.1.7c

Bug fixes:

• Changes XS:i field to ZS:i field as in HISAT so it doesn't conflict with XS:A field. This issue had been affecting compatibility with Cufflinks/StringTie.
• Comment lines in the manifest file no longer crash preprocessing.

Rail-RNA v0.1.7b

Performance enhancement:

• By default, Rail-RNA no longer attempts to align reads with long poly(A) tails that, when trimmed, are smaller than --min-exon-size. This can significantly improve load balance during the first alignment step.

Rail-RNA v0.1.7a

Critical fix:

• Rail-RNA no longer skips second-pass alignment in its first alignment step; in v0.1.7, it did, disappearing some reads during the pipeline. Do not use v0.1.7.

Bug fixes:

• Rail-RNA is now forgiving of gzipped input files that end unexpectedly.
• --skip-bad-records is now more robust. Previously, it would choke on bad FASTQ records with missing lines.
• FASTQ downloads from remote URLs now work in local mode.

Rail-RNA v0.1.7

Because of a bug in align_reads.py where the script is terminated early and final output is missing reads, this version is deprecated.

Bug fixes:

• The first half of the pipeline (through building the transcript index) no longer hangs on shuffle/sort for thousands of samples in elastic mode. Previously, too many reduce tasks were being written in each step.
• The readlet alignment step no longer has a significant memory footprint when a single readlet is found in many reads. Previously, read information for all reads with the same given readlet was loaded into memory at once.

Performance improvements:

• The speed of the first alignment step has been nearly doubled for pipelines where the only deliverable is an isofrag index.
• By default, readlets spanning poly(A) tails are now eliminated before aligning readlets. This significantly improves load balance when operating on samples where poly(A) tails have not been trimmed.

Rail-RNA v0.1.6c

Bug fixes:

• Previously, reference names with parentheses returned an error. References names with nonalphanumeric characters are now accommodated.
• In elastic mode, if Rail-RNA creates a bucket on behalf of the user, that bucket will be in the same region as the region to which a job flow is being submitted. Previously, the bucket would always be in us-east-1.

Rail-RNA v0.1.6b

Bug fixes:

• The intron search step no longer returns an error when it tries to search for splice sites beyond the extent of a reference contig.
• The minimum exon size (--min-exon-size) is now respected when enumerating configurations of introns overlapped by transcript fragments.

Rail-RNA v0.1.6a

Bug fixes:

• Solexa-encoded quality strings are now recognized and interpreted properly. Previously, they were misidentified as either Phred64- or Sanger-encoded.
• Slow shuffle-sort in the first alignment step at scale in elastic mode has been mitigated by halving the number of concurrently executing reduce tasks.
• Preprocess job is not terminated prematurely in elastic mode.