Changes:

Novoalign

1. Fix: Previous V3.06 releases may underestimate MAPQ for some reads.
2. Fix: Novoalign (V3.06.03) output an invalid SAM record if input read was zero length

Novosort

1. Fix: V1.04.02 wrote invalid BAM headers to the temporary files and failed if sort did not fit in RAM

Refer to readme.txt for a full list of changes.

Changes:

Novoalign

1. Fix: Adapter trimming of single end reads (-a option) can work inconsistently if reads have already been adapter trimmed by a pre-process.

Novosort

Fix: Novosort could use excessive memory if BAM file had a huge number of @SQ records. This change frees @SQ memory before allocating the sort buffers.

Novosort in this release is broken and will fail if sort is not memory contained.

Refer to readme.txt for a full list of changes.

Changes:

1. Novoalign* – Fix: Novoalign may fail to detect truncated block compressed read files and proceed as if EOF was reached.
2. Novoalign[CS]MPI was dynamically linked since V3.05.00. It is now statically linked.

Refer to readme.txt for a full list of changes.

Changes:

Novoalign

1. Fix: Novoalign’s –amplicon option was soft clipping one too many bases from 3′ ends of targets due to interpretation of BED file ThickEnd as zero based coordinate of last  thick base when it should be interpreted as the base after the last thick end base.
2. The softclipping algorithm has been changed to use the –matchreward setting. Previously it used a matchreward of 8 while the –matchreward setting was used in earlier alignment stages that use Glocal Needleman-Wunsch algorithm. Setting –matchreward 3 is useful to soft clip ragged ends from NextSeq alignments.

Refer to readme.txt for a full list of changes.

Changes:

Novoalign

1. Added optional SAM tags for mismatches(XM:i) and gap opens (XO:i). To enable the tags add option –tags XM XO to the command line.
2. Performance tuning. In V3.02.7 we removed the 15bp maximum indel allowed in the anchor read of a pair to better handle long indels in paired reads that overlap (short fragments). We have changed this so that we first check for an overlap and set maximum indel length to 50% of overlap or 15bp, which ever is larger. This has restored runtime performance to V3.02.06 levels for reads with minimal overlap.
   Tests on variant calling have shown no negative affects.

Novoutil

1. Added a function, man2bed, to convert Illumina manifest files into bed files for use with novoaligns –amplicons option.
   Usage: novoutil man2bed manifestfilenameDescription:
   Illumina amplicon manifest file is processed to create an amplicon bed file for novoaligns –amplicons option and a target region bed file that can be used for filtering vcf files. The [Probes] section of manifest must be before [Targets] section.Output bed file names are created from the manifest file name by removing any ‘.txt’ suffix and then adding ‘.amplicons.bed’ or ‘.regions.bed’.

Refer to readme.txt for a full list of changes.

Changes

1. All Programs have been updated to accept –help option.
2. When printing command line arguments to log file or SAM @PG record
   a) Zero length arguments are printed as “”
   b) Any tabs embedded in arguments are printed as \t
   c) Any other control characters are printed as octal \o999
3. Fix: In V3.05.00 running novoalign with no arguments would result in a seg fault.

Changes

Novoalign*

1. Fix: Novoalign may produce alignments that can’t be sorted with novosort –markduplicates due to a SAM formatting error on some chimeric pairs.
2. Fix: When using –alt option multi-mapped alignments may not be filtered using -R limit. This can mean excess alignments are reported when using -r All. When not using -r All the NH,IH & HI tags may be set to show a multi-mapped alignment even though the best alignment has a high MAPQ.
3. Fix: Novoalign could enter a CPU loop if one read of pair had just enough bases to pass QC checks (at default settings) and the other read failed.

Refer to readme.txt for a full list of changes.

Changes:

1. Fix: NovoalignMPI may hang if Illumina chastity filter flag is set to Y on a read.
2. Fix: Novoalign can produce invalid CIGARs for multi-mapped reads when using -r Random or -r All (from V3.04.00)

Changes:

1. Added an option to transfer FASTA/Q header comments to back end of SAM report lines as per bwa mem’s -C option.
2. Fix: Novoalign Seg Faults if read file is empty. (from V3.03.00)
3. Fix: Memory leak when input read files were block compressed (bgzf)
4. Fix: NovoalignMPI may seg fault if Illumina chastity filter flag is set to Y on a read.

Refer to readme.txt for a full list of changes.

Changes:

Novoalign[CS]

1. Fix: For pairs with one read unmapped the SAM flags may be set incorrectly in the unmapped read (Mate not Mapped set) causing novosort markduplicates to report an error.

Refer to readme.txt for a full list of changes.

Changes:

1. Fix: Changes in V3.03.00 meant Novoalign would only process the first block of block compressed read files.

*** This release has issues reading block compressed gzip files ***

Changes:

Novoalign

1. Fix: Novoalign slows down if a significant proportion of paired end reads have fragment lengths shorter than the read length and adapter trimmming is on. This change aligns one end in single end mode and then maps the mate to the same location. There is a small reduction in number of proper pairs as the change applies -t setting individually to each read of a short fragment rather than to the pair.

NovoLR

1. Fix: novolrcleaver would not split records if the header contained white space.
2. Add option -SVSplit to novolrcorrector that splits reads at locations not covered by proper pair alignments
3. Run report now includes counts of corrected indels by length
4. Fix: -uncorrectedLR option was not outputting trailing long reads with no short read mappings.

Refer to readme.txt for a full list of changes.

*** This release has issues reading block compressed gzip files ***

Changes:

1. Added extra threads for inflating read files and building MPI messages. This allows MPI programs to effectively use more nodes.
2. Added a Match Reward option to Novoalign[MPI]. Setting option –matchreward 3 improves mapping of deletions near the ends of the reads. Values around half the gap extend penalty work well.
3. Initial release of NovoLR programs for correcting single molecule long reads using mapped Illumina reads.

Refer to readme.txt for a full list of changes.

Changes:

1. Novomethyl. Changes to calculation of conversion efficiency. Previous calculation used ratio of T/(C+T) calls for all CHG & CHH sites. We now report two estimates, the first based on T/(C+T) ratio for cytosines called as unmethylated with a quality > 30 and a read depth (C+T) > 10, the second uses T/(C+T) ratio for all cytosine sites and is useful for spike in samples where it is known that all cytosines are unmethylated.
2. Novosort. Fix: Novosort can stop with “Mate Read not found” when marking duplicates on BWA alignments when one read of a pair is unmapped and mappings have inconsistent strand flags.

Refer to readme.txt for a full list of changes.

Changes:

1. Novoalign Fix:  If an Illumina read has failed the chastity filter novoalign attempts to write a zero length line which results in a write failure.
2. Novosort  Fix: When merging and marking duplicates on BAMs which have already had duplicates marked then the existing duplicate flags are not reset. This is not a problem if duplicates are chosen in the same way but this is only likely to be the case if  using novosort –markduplicates for the original and the merge and the –keeptags option was used.

Changes:

Fix: When sorting and marking duplicates on a BWA MEM alignment file novosort may stop with error “Mate Read not found”.  This happens when BWA MEM reports a secondary alignment for a pair where mate is not  mapped and the secondary alignment RNEXT & PNEXT are set to secondary alignment location rather than the primary alignment location as recommended in the SAM specifications.

Changes:
Novosort

1. If –markduplicates is used on a pre-sorted paired end BAM file that doesn’t have the Z5 & ZQ tags then report an error and stop. Previous versions dropped back to a samtools rmdup like mode.
2. Moved buffer allocation code so that no buffers are created if all input files
   are already coordinate sorted.
3. Fix: If a BAM file had > 1Gbyte of headers then novosort would crash or hang.
4. Fix: If [–index|-i] option is used with –namesort then novosort still attempts
   to index the BAM file resulting in huge index and possibly memory allocation failure.
   The fix disables index creation for name sorted BAM files.

Changes:
Novosort

1. Fix: Novosort would crash with a seg fault if a read flagged as mapped had BAM alignmnet attribute refID = -1

Novoalign

1. Fix: When using -H to hard clip 3′ low quality bases, if all bases would be clipped by this rule then the read was left intact (no clipping was done). This meant novoalign would attempt to align the read with low qualities. This was not an issue if all base qualities were 2 ‘#’ but is a problem if all base qualities are around 10 and we are using -H 15 or similar to trim them.
2. The -H option trims 3′ low quality bases with a modified Mott’s algorithm. One parameter of this algorithm was hidden in previous versions and is now exposed. Please refer to the manual.
3. Added an option to write an Amplicon BED file with score column equal to count of read pairs aligned to the amplicon.
4. Fix: -p option for filtering polyclonal reads was not working to specification. In Novoalign the second pair of values specified the number of bases below the threshold rather than the fraction of bases. This fix changes Novoalign to work the same way as NovoalignCS and use the fraction of bases.

Novoalign[CS]MPI

1. In the master processes messaging code MPI_Waitany was changed to MPI_TestAny with a nanosleep() if no message was received. This reduces CPU time for the master process allowing more CPU for slaves on the same node. Thruput increases by about 50% of 1 CPU core.

Refer to readme.txt for a full list of changes.

Changes:

1. Fix: Counts of soft clipped amplicons were wrong if there were more than 1000 amplicons.
2. Fix: File system write status was not checked on some writes resulting in normal program termination when in fact the output file was broken. (e.g. Not Enough Space on Device errors)
3. Soft clipping of amplicon primers now allows a user specified deviation in primer alignment location –amplicons <bedfile> [delta]. This allows the fragment mapping  to start ‘delta’ bp outside the amplicon coordinates specified in the BED file.
4. Added option –hugePage to allocate index using huge pages. Note. –mmapoff will use anonymous huge pages if available. In tests, 1G huge pages gives >5% performance improvement.
5.  Fix: NovoalignMPI Seg Faults when run with quality calibration and Single End reads.

Novoalign

1. Fix: Using quality calibration reruns were not concordant and slight differences could be seen in calibrated base qualities for some runs. This may also have caused slight differences in alignment quality between reruns.
2. Added a new option for processing a subset of reads. -# X:Y skips X reads and then processes every Yth read.
   Examples..
   -# 0:10 will process reads 1,11,21, etc.
   -# 1:10 will process reads 2,12,22,…
   -# 1M -# 100000:1 will skip 100000 reads and then process every read until 1M reads have been processed.
3. Fix: –hlimit option does not work with paired end reads
4. Base quality calibration has been disabled for BiSeq alignments as it can cause quality of cytosine bases to drop sufficiently to adversely affect methylation calling.

Novoutil IUPAC

1. Now adjusts thickStart & thickEnd when lifting over a BED file.

Refer to readme.txt for a full list of changes.

Changes

Novoalign

1. Fix: In BiSeq mode, if the reference genome was indexed with a kmer >16bp the seeding process would  still use  16bp seeds  resulting in seeding  to an excessive number of locations. The fix improves performance of BiSeq alignments.
2. Fix: In BiSeq mode memory usage could still grow. The fixes reduces memory and reduces runtime. Some high aligment score paired alignmenst may now be come back as unmapped.These would have had an alignment score in excess of the -t setting in previous versionsand were possibly false positive laignments.
3. Fix: Assertion `tgtlen <= ncols – 32′ failed. This could happen when using -a optionin paired end mode when mean fragment length <= read length.
4. Fix: Novoalign crashes if BED file for amplicons contains a comment line.

Novosort

1. Check that the temporary folder is writeable during initialisation. In previous  versions if temporary folder didn’t exist novosort would fail when it first tried  to create a temporary file or complete successfully if temporary files were not required.

Refer to readme.txt for a full list of changes.

Changes:

1. Novoalign: Fix: Memory usage increases until we get an allocation failure. This was more likely
   to happen on Bi-Seq alignments.
2. Novosort: Fix: Novosort was stopping with “Invalid BAM Header” when processing BAM files from ISAAC.
3. Novoutil: IUPAC BED file relocation now relocates thickStart & thickEnd.

Refer to readme.txt for a full list of changes.

Changes:

1. Novoalign – Earlier versions of Novoalign had a 15bp indel limitation on the read that anchors a pair alignment. This limitation was causing problems with the alignment of long reads that have large overlaps such that both reads of the pair have the same long  indel. This change removes the15bp indel limit.

Major Changes:

1. Fix: Novosort –markduplicates -i creates an invalid BAM index.

Refer to readme.txt for a full list of changes.