Connor 0.6.1

Command-line tool to deduplicate reads in bam files based on custom inline barcoding.

A command-line tool to deduplicate bam files based on custom, inline barcoding.

The official repository is at: https://github.com/umich-brcf-bioinf/Connor

Overview

When analyzing deep-sequence NGS data it is sometimes difficult to distinguish sequencing and PCR errors from rare variants; as a result some variants may be missed and some will be identified with an inaccurate variant frequency. To address this, researchers can attach random barcode sequences during sample preparation. Upon sequencing, the barcodes act as a signature to trace the set of PCR amplified molecules back to the original biological molecules of interest thereby differentiating rare variants in the original molecule from errors introduced downstream.

Connor accepts a barcoded, paired alignment file (BAM), groups those input alignments into families, combines each family into a consensus alignment, and emits the set of deduplicated, consensus alignments (BAM).

Connor workflow:

Sequencing [FASTQ 1/2] -> Aligner [BAM] -> Connor [BAM] -> Variant Detection [VCF]

Connor first groups original alignments into alignment families based on their alignment position and Universal Molecular Tag (UMT) barcode. (Connor assumes the incoming aligned sequences begin with the UMT barcode.) Each family of alignments is then combined into a single consensus alignment; discrepancies in base-calls and qualities are resolved by majority vote across family members. By default, smaller families (<3 align pairs) are excluded.

For more information see:

• QUICKSTART : get started deduplicating barcoded BAMs.

• INSTALL : alternative ways to install.

• METHODS : details on UMT barcode structure, suggestions on alignment parameters, details on family grouping, and examples.

Connor help

$ connor --help
 usage: connor input_bam output_bam

 positional arguments:
   input_bam             path to input BAM
   output_bam            path to deduplicated output BAM

 optional arguments:
   -h, --help            show this help message and exit
   -V, --version         show program's version number and exit
   -v, --verbose         print all log messages to console
   --log_file LOG_FILE   ={output_filename}.log. Path to verbose log file
   --annotated_output_bam ANNOTATED_OUTPUT_BAM
                         path to output BAM containing all original aligns annotated with BAM tags
   -f CONSENSUS_FREQ_THRESHOLD, --consensus_freq_threshold CONSENSUS_FREQ_THRESHOLD
                         =0.6 (0..1.0): Ambiguous base calls at a specific position in a family are
                          transformed to either majority base call, or N if the majority percentage
                          is below this threshold. (Higher threshold results in more Ns in
                          consensus.)
   -s MIN_FAMILY_SIZE_THRESHOLD, --min_family_size_threshold MIN_FAMILY_SIZE_THRESHOLD
                         =3 (>=0): families with count of original reads < threshold are excluded
                          from the deduplicated output. (Higher threshold is more
                          stringent.)
   -d UMT_DISTANCE_THRESHOLD, --umt_distance_threshold UMT_DISTANCE_THRESHOLD
                         =1 (>=0); UMTs equal to or closer than this Hamming distance will be
                          combined into a single family. Lower threshold make more families with more
                          consistent UMTs; 0 implies UMT must match
                          exactly.
   --filter_order {count,name}
                        =count; determines how filters will be ordered in the log
                        results
   --umt_length UMT_LENGTH
                        =6 (>=1); length of UMT

---

Email [e-mail] for support and questions.

UM BRCF Bioinformatics Core

Changelog

0.6.1 (8/16/2018)

• Adjust logging to not crash if username/hostname not available

0.6 (4/11/2018)

• Extended to support pysam v0.13, v0.14

• Added optional command line arg to specify length of unique molecular tag (UMT)

• Added optional command line arg to sort filters results by name instead of count

• Added validation to check for properly paired alignments

• Added validation to check for presence of secondary alignments

• Adjusted so warning instead of error when no families found

• Substantial refactors to clarify implementation

0.5.1 (9/8/2017)

• Extended supported python and pysam versions

• Adjusted to avoid performance problem when processing extremely deep pileups

• Adjusted so that when no families pass filters show warning instead of error message (thanks to ccario83 for upvoting this fix)

0.5 (9/13/2016)

• Filters now exclude supplemental alignments

• Added BAM tags to show pair positions and CIGAR values

• Reduced required memory and improved performance

0.4 (8/26/2016)

• Added input/command validations

• Added annotated bam option

• Revised QUICKSTART, METHODS

• Added PG line in BAM header

• Improved logging of filtered aligns and progress

• Removed some logged stats to focus logging results

• Removed dependency on pandas/numpy

• Moderate performance (speed) improvements in calculating consensus sequence

• Switched consensus quality to be the max mapping quality

0.3 (8/8/2016)

• Added filters to exclude low quality, unmapped, or unpaired alignments

• Revised BAM tags; documented BAM tags in BAM header

• Extended logging to write to file and console

• Adjusted to make deterministic in Py3/Py2

0.2 (7/15/2016)

• Bugfix: connor was mangling left hand side of right hand consensus reads

• Fuzzy grouping of pairs into families based on left or right UMI match

• Fuzzy grouping of pairs into families based on UMI within Hamming distance

• Command line args for hamming distince, consensus threshold, min orig reads

• Extended logging to assist in overall diagnostics

• Generate additional file of alignments excluded from consensus (diagnostic)

• Added UMI sequence tag (X0)

0.1 (6/17/2016)

• Initial development release

• Partitions raw reads into consensus families

Connor is written and maintained by the University of Michigan BRCF Bioinformatic Core; individual contributors include:

• Chris Gates

• Peter Ulintz