Novoalign & MPI

1. Fix: Novoalign could generate a bad alignment if chrM was the first sequence in
   the genome and reads crossed the end points of the circular chromosome. Most
   alignments were correct but occasionally an alignment would show as an insert
   before the chrM sequence generating a bad RNAME

Novosort

1. Include soft clipped bases in the callable regions bed file generation

Novoutil

No change

Novoindex

No change

Files

All programs are statically linked and should run on any X6-64 Linux system

All

1. Change compression library from zlib to zlib-ng for Centos 6 build. This improves performance of all programs
   that inflate or deflate *.gz or *.bam files.
   In Novosort CPU is reduced about 40% and wall time about 10%.

Novoalign

1. Fix: Novoalign was writing an SA:Z: even tag when –tags -SA was used.
2. Fix: If an alt_scaffold sequence header used a “:” separator rather than
   a ‘-‘ bewteen the from & to positions they would be parsed incorrectly.

Novoutil

Novoutil readcover

1. Fix: Seg Fault could occur due to access passed end of histogram data by 1
2. Add a MAPQ limit to reads counted as cover. This allows for case that very low MAPQ alignments are not useful for variant calling.

Novoutil kmerfilter

1. Added multithreading for improved performance.

Novosort

1. Fix: Novosort could hang while marking duplicates if there was a very high
   read depth. This occurred on one known dataset near centromere on chr1.
2. Fix: Novosort callable regions bed file was missing some regions that were callable.
3. Fix: Novosort could output a zero length bed file entry for the callable bed file if
   the end of a read reduced cover below the high cover limit and another read started
   at the same locus pushed the cover back over the limit. This should not be a problem
   if you use the recommendation of adding a extra base pairs to each callable region using
   novoutil splitbed.
4. Name sorting now puts secondary alignments before supplementary alignments. In previous
   releases the order was undefined.

Files

Novoalign

1. Fix: In BAM output format the optional tags ‘Ms’ & ‘Ps’ were not being added even though
   command line options requested them.
2. Fixed multiple compiler warnings after upgrade to latest GCC compiler.

Novoutil

1. Added function readcover to produce a histogram of read depth for regions in a bed file.

Files

Novoalign

1. Fix – When reporting alignments to alt_scaffolds reporting would stop after 1000 equal best
   alignments for a read even if the read mapped to more than 1000 alt_scaffold sequences. This was
   an issue in an experiment where we had large numbers of alt_scaffolds for highly variable regions
   of the genome. The problem is not seen when using GRCH37 & 38 alt_scaffolds.
2. Fix – When reporting alignments to alt_scaffolds reporting could stop early if a read mapped
   with equal best alignment score to multiple alt_scaffold regions. This was first seen when
   mapping single end RNA reads to a reference with a large number of HRV regions and associated
   haplotypes (alt_scaffold sequences). The problem is unlikely, but possible, when using paired
   end reads with GRCH37&8 alt_scaffolds.

Novosort

1. Rebuild with no changes.

Novoutil

1. kmerfilter function can now filter fastq files.

Novoindex

1. Rebuild with no changes.

Files

Novoalign

1. Fix: Use of c++ alignas(32) does not work for some combinations of Linux & gcc versions.
   We changed all instances to use posix_memalign()
2. Changes to how tags on FASTA headers of the reference are copied to the @SQ records to
   support user defined lower case tags. Now all tags in form “[a-z][a-z]:.*\t” are copied
   to the @SQ records. Other tags that are copied or renamed are…

In previous releases the following tags were copied and renamed…

NovoalignMPI

1. (see Novoalign)

Novosort, Novoutil, Novoindex

1. Rebuild with no changes.

Files

Novoalign

1. Rebuild due to changes in libraries common with Novosort.

NovoalignMPI

1. Rebuild due to changes in libraries common with Novosort.

Novosort

1. Fix: Novosort consensus read generation has an option to allow unique reads to be included in the
   consensus file however the XZ tag (number of reads used to generate the consensus read)for
   unique reads was set to 0(zero) rather than 1.
2. Fix: Marking duplicates on Single end reads uses excessive memory.

Novoutil

1. novoutil hlaextract can now take a bed file to identify regions to extract.
2. novoutil kmerfilter is now based on % of bases covered by k-mers rather than the number of matching kmers.

Novoindex

1. No Change

Files

Novoalign

1. Minor changes to –help text
2. Rebuild due to changes in libraries common with Novosort.

Novosort

1. Minor changes to –help text.
2. Addition of csi index creation. The index will automatically switch to csi format if any reference sequence is over the length limit for bai indexes. You can also force csi index format by using the option –csi rather than -i.Novoutil

Novoutil

1. Fix: novoutil groupbyname was doing a double free which could cause a seg fault.
2. Improve novoutil groupbyname –help text
3. Improve novoutil hlaextract –help text.
4. Improve novoutil kmerfilter –help text.

Files

Novoalign

1. When using –amplicons option Novoalign now checks maximum possible length of a primer dimer
   and may suggest using the -l option to avoid mapping them. If mapped they can result in fully
   soft clipped alignments which may break GATK tools.

Novosort

1. Fix: Novosort consensus reads change the readgroup tag on the consensus reads by adding a suffix
   of ‘.NSC’ but the @RG header wasn’t updated. We now add the suffix to @RG ID: tag as well.
   Users can set the suffix with the –cns option.

Novoutil

Added several new functions

1. hlaextract – Extracts primary alignments mapped to chr6 MHC region and unmapped reads from a bam file. The output is grouped by name.
2. kmerfilter – Matches reads against a fasta file using 16 mer counting and can remove (or keep) alignments with less than some % of matching 16 mers.
3. groupbyname – Given a BAM input it groups reads by readname for use in realigning. Secondary, QC’d and Supplementary alignments are removed.

Files

Novosort

1. Fix: When writing singleton reads (–consensus with minReads set to 1) to the consensus read file the SEQ & QUAL fields of -ve strand alignments were not reverse complemented.
2. Fix: Secondary & Supplementary alignments were not being marked as duplicates or removed from the ouput BAM even though a consensus read had been generated from the primary alignments.
3. Fix. UMI correction failed when using a valid UMI list, producing a badly formatted bam file.
4. Changed the log to report alignment counts rather than pairs. This is necessary as some secondary and supplementary alignments may not be paired. (this was done in Novosort V3.01.04 but not officially released)

Files

Novoalign

1.  Add ‘+-‘ option to 5′ UMI extraction. When used with UMIs on both ends of a fragment it puts the most 5’ UMI first in the SAM tag rather than the read1 UMI first such that reads from the two original strands end up with the same UMI tag. This allows Novosort to recognise duplicates mapped to opposite strands and to produce a single consensus record from both strands.
2. Fix: If processing reads from an unmapped BAM file and using -# option to limit the number of reads Novoalign could hang after processing the requested number of reads.

Novosort

1. Revised the method for calculating consensus base quality. It now uses frequentist inference.
2. Fix: Gradual growth in memory use due to temporary file segments not reusing deflator buffers.
3. Fix: Name sort in V3 was broken.
4. Add Supplementary flag to name sort key so that supplementaries come after primary alignments.

Files

novoSplice v0.8.4 (11th December 2020)

Minor release improving parsing options, help-screen and info.txt output file
Changes log since v0.8.3:
Bugs:

1. [SPLICE-332] – Fix auxiliary info.txt file
2. [SPLICE-334] – Fix output files prefix
3. [SPLICE-335] – Fix –tune option prevents changing user defined options

Files

Novosort

1. Since V3.00 release, duplicate detection for single end reads was not functioning as designed. Single end
   duplicate testing was on by default and the –includeSE option was turning it off. There was also a
   failure to count single end reads when the option was off.
2. Also since V3.00 release, Default optical duplicate distance was not being set leading to optical duplicates being counted as
   “Excess Same Tile” duplicates. It did not affect which alignments were flagged as duplicates.

Novoalign

1. Added option –SuppNoSQ, for supplementary SAM records put ‘*’ in SEQ and QUAL fields. This is similar
   to the –2NoSQ option for secondary alignments.
   We are using this when genotyping alternate scaffolds based on alignment cover. It shouldn’t be
   used if you intended to call variants from the supplementary alignments.

Files

Novosplice

1. Initial public release of our RNA aligner. Novosplice can handle short and long reads as from Illumina, BGI, Pacbio and Nanopore.

Files

Novoalign

1. Approx 30% performance improvement for mate pair alignments when using long & short fragment length option -i MP 99[-|,]99 99[-|,]99

Files

This release is focussed on adding functionality to Novoalign and Novosort for handling UMIs, MBCs and generating consensus reads from duplicate groups. Other changes include improved handling of @PG and @CO records and a couple of fixes for assert failures when processing mate pair alignments.

Novoalign

1. When extracting UMIs from both read 1 & 2 (-5 option) add a hyphen delimiter between the UMIs in the RX & QX tags.
2. The -5 option now allows extraction of both cellular barcodes (CR tag) and UMIs (RX tag) from 5′ of a read.
3. New –umi option allows input of an UMI only read file with UMI extraction similar to -5 option.
4. New option –umi2 to extract UMIs from read header and store into a SAM tag.
5. When mapping an (unaligned) BAM file any @PG or @CO records are now copied to the output BAM file. Previously
   only @RG was copied.
6. Fix: Assert failure in rptCIGAR() method
7. Implement a –slidingWindow base quality trim option as alternative to -H option.
8. Fix: Mate pair alignments could get an assert failure in method findmate256()

Novosort

1. UMI aware mark duplicates now allows inexact matching and correction of UMIs and provision for whitelists
   and clustering of up to 3 UMI/Barcode tags. Functionally similar to UMI-Tools
   Novosort duplicate marking should be concordant on reruns. Other python based tools may not be.
2. Fix: Novosort could add a second set of Z5 & Zq tags if an unsorted input file already had
   these tags.
3. Novosorts ranking of duplicates uses the sum of all base qualities to determine highest
   quality read of a signature group. Samtools markdup only sums base qualities > 15.
   We’ve added option –minBaseQuality so you can set a lower limit on base quality for
   ranking. Set –minBaseQuality 15 to get samtools compatible ranking and an increase in CPU time.
4. Improved multi-threading and other performance improvements can give ~33% elapsed time
   reduction when Marking Duplicates. The actual % drop you’ll see depends on how many
   cores you are using and how fast your IO system is.
5. New option –consensus for generating consensus reads from duplicate groups. Functionally
   similar to fgbio, http://fulcrumgenomics.github.io/fgbio/tools/latest/CallMolecularConsensusReads.html.
6. Addition of –stats option to write tsv format file with duplicate statistics.
7. Novosort now adds it’s own @PG record and updates PG tag of any SAM records that it updates. Records
   that only had the order changed keep their original PG tag.

Files

Novoalign

1. Fix assert failure in CIGAR generation routine.

Files

Novoalign

1. Fix: Assert failure in rptCIGAR() method

Files

Novoalign & NovoalignMPI

1. Reruns of Novoalign V4 could be discordant. Differences were seen in reads
   multi-mapped to repeats with very low alignment scores and
   with MAPQ <= 1 and affected the AM & SM single end quality tags but could
   also change MAPQ between 0&1.
2. Fix: When using –tune HiSeqX the polyclonal settings were misreported, showing as -p 1,4 when it was -p 4,20
3. Fix: When using –tags MC the Mate CIGAR tag was not being added to chimeric supplementary alignments.

Files

Novoalign

1. Add option to disable generation of M5 checksums for @SQ records.
   –addM5 [on|off] Default is On.

Novoindex

1. Adjust default index k&s for better performance on Human reference genome.
   Default now builds a k=15, s=2 index rather than a 14/2.

Novoutil

1. Add function ‘addFastaM5’ to add M5 and LN metadata tags to fasta headers.

Files

Novoalign

1. Fix: Novoalign paired end chimera mode could produce invalid alignments
   across circular chromosome junctions if read sequence had N’s at the junction
2. Fix: novoalign: src/rptSAM.cc:775: void rptSAM::rptcigar(): Assertion `tgt[tposn] != ‘\0” failed.

Novosort

1. Performance tuning of mark duplicates code. This should reduce run time on high core count servers.

Novoutil

1. Fix: Splitbed function could produce bed file with coordinates outside chromosome size when
   using the –add option. To fix we now have an option (–sam) to input a SAM format header
   file and then extract sequence lengths from the @SQ records.

Files

Novoalign

1. In base quality calibration (-k) limit maximum base quality to 60. (GATK limit)
2. Fix: When compiling with gcc 7.4, BAM output format shifted base qualities one to the right. This only
   happened on our development system after upgrading gcc and isn’t in any official releases.
3. Increased sensitivity of chimeric read mapping by removing structural variation penalty from calculation.

Novosort

1. Add option –callableRegions to write a bed file of callable regions. “novosort –help” for more details.

Novoutil

1. Added function splitbed that divides a bed file into multiple files of approximate equal size. We have used this with WGS
   sequencing and novosort –callableregions to create per core  bed files for variant calling that are equal size.
   For WES we use target capture region bed files as input rather than the callable regions file from Novosort
   “novoalign splitbed –help” for details

Files

Novoalign

1. Fix: When using alternate scaffolds the mate CIGAR tag (enable with –tags MC ) could have the CIGAR of the last supplementary alignment rather than the primary alignment.

Files

Novoalign

1. Fix: If Novoalign is given empty input file it wasn’t writing BAM headers and exited with an error status. This change will produce a valid SAM file with headers only and exit with normal status.
2. Fix: When several alt-scaffold alignments have the same best alignment score (and better than main chromosome alignments) the primary alignment was not selected randomly.
3. Fix: Alignments on systems with AVX2 were not concordant with systems without AVX2. This affected MAPQ calculation and also random selection of low MAPQ alignments.

Files