ProDuSe2

Variant caller that utilises semi-degenerate barcoded adapter libraries.

To install: pip install produse Or pull from the official repo:

git clone [email protected]:morinlab/ProDuSe.git
cd ProDuSe
python3 setup.py install

To use:

$ produse
Usage:   produse <command> [options]
Commands:
adapter_predict Identifies the degenerate barcode used in a set of FASTQ files
run             Runs all steps of the ProDuSe pipeline listed below
trim            Removes barcodes from FASTQ files
collapse        Identifies duplicates,and merges duplicates into a consensus
call            Identifies variants
All commands accept -h for their specific usage.

The relevant publications can be found here:

Note:

• ProDuSe makes use of ClipOverlap in its pipeline with arguments -abt8.
• bwa mem is the aligner used.

Collapse

Collapse reduces records with the same reference start position, forward/reverse flag and barcode within a specified Hamming distance to a single concensus family record. A consensus family record is essentially a pileup of the family member records and the most common operation/base at each position is retained. The families are first collapsed if the barcodes are identical into "proto-families". These proto-families barcodes are then compared to each other producing a graph with node weights of the proto-family size and edge weights of their Hamming distances. The graph is reduced to a minimum spanning forest where any edges above threshold are removed. Each tree in the forest is further reduced to a diameter twice the threshold while maximising each subtrees node weight. A concensus of each tree is then finalised.

Each family record is output with two additional tags:

• fQ: Integer array containing Phred score of wrong base chosen during collapse
• fC: Integer array containing pairs (simmilar to a CIGAR) of count and depth representing, from the start of the alignment, the depth of the family at a position

Call

Configuration

While all arguments can be passed via command line, the main pipeline requires so many that a config file is recommended. Config files are in YAML format. See configutator for more details.

Here is an example config.yaml with all parameters specified:

default: &default
  output: ./
  reference: Ref.fa
  bwa: bwa mem -t 10
  samtools: samtools

  trim: &trim_default
    barcode_distance: 3

  collapse: &collapse_default
    barcode_mask: "SSS1111111111110"
    barcode_distance: 3

  filter: &filter_default
    min_molecules: 2
    mutant_molecules: 3
    variant_allele_fraction_threshold: 0.01
    min_reads_per_uid: 1

samples:
  - name: test1
    <<: *default     # Inherit default parameters
    fastqs:
      - test/test1/onm_R1.fastq
      - test/test1/onm_R1.fastq
    trim:
      <<: *trim_default
      barcode_sequence: NNNWSMRWSYWKMWWT

  - name: test2
    <<: *default     # Inherit default parameters
    fastqs:
      - test/test2/afnm_R1.fastq
      - test/test2/afnm_R2.fastq
    trim:
      <<: *trim_default
      barcode_sequence: NNNWSMRWSYWKMWWT

As you can see, multiple samples can be specified in the same config file and will all be processed with a single run of ProDuSe.