config.yml

############################
##### INPUT PARAMETERS #####
############################
# The rundir parameter is the name of a subdirectory under data/ that must contain
# - asv_seqs.fasta (ASV sequences in FASTA format) 
# - asv_counts.tsv (counts of ASVs (rows) in samples (columns)) 
rundir: "test"
# The run_name parameter is the name of the workflow run and is used to separate
# runs with specific parameters of the clustering tools. 
# With different run_name parameters on the same rundir you can cluster the ASVs
# while automatically using the same alignment output (from vsearch).
run_name: "run1"
# With the split_rank parameter you can split the ASVs by a certain taxonomic rank
# prior to alignment and clustering, which parallellizes the workflow and can help
# speed things up on large datasets. 
# Note that the split_rank must exactly match one of the columns in the asv_taxa.tsv file
split_rank: "Family"
  
####################################
##### PREPROCESSING PARAMETERS #####
####################################
preprocessing:
  # Whether to perform length filtering of input seqs
  filter_length: False
  # Minimum and maximum length of input seqs
  min_length: 403
  max_length: 418
  # Whether to perform codon filtering of input seqs
  filter_codons: False
  # Specific stop codons to look for. Specify multiple codons with in a comma separated list like: "TAA,TAG". By default TAA and TAG are being looked for.
  stop_codons: "TAA,TAG"
  # Specific position of the ASV where to start checking for codons.
  start_position: 2
  # Specific position of the ASV where to stop checking for codons. By default it checks until the end of the sequence
  end_position: 0

###############################
##### TAXONOMY PARAMETERS #####
###############################
# The split_size parameter specifies the number of sequences to include in each
# split fasta file. This is used to reduce the size of the fasta files that are
# input to the taxonomy assignment tools.
split_size: 500
# The taxonomy_source parameter specifies what taxonomy source to use for the
# clustering. This is the primary source of taxonomy information and is used to
# to partition the ASVs into smaller fasta files based on the split_rank
# parameter. If you already have a tab-separated file with taxonomic
 # assignments, set taxonomy_source to the path of that file.
taxonomy_source: "sintax+epa-ng"
# The ranks parameter specifies what taxonomic ranks are included in the optional
# ASV taxonomy file. If you don't have a taxonomy file, and instead want to assign 
# taxonomy using the taxonomic tools specified under 'taxtools', do not modify this
# parameter. If you have a taxonomy file, set this parameter to the ranks that are 
# included in the file, e.g. ["Kingdom", "Phylum", "Class", "Order", "Family", "Genus", 
# "Species", "BOLD_bin"]. If you leave this parameter empty the pipeline will attempt to
# guess which ranks are included in the taxonomy file.
ranks: []
# The taxtools parameter specifies what taxonomic tools to use for assigning taxonomy
taxtools:
  - "sintax"
  - "vsearch"
  - "epa-ng"
  - "sintax+epa-ng"
# The sintax parameter specifies the parameters to use for the sintax taxonomy
# assignment tool.
sintax:
  ref: ""
  randseed: 15
  cutoff: 0.8
  ranks: 
    - "Kingdom"
    - "Phylum"
    - "Class"
    - "Order"
    - "Family"
    - "Genus"
    - "Species"
    - "BOLD_bin"
# Below are the parameters for qiime2 (vsearch and sklearn) taxonomy assignment
qiime2:
  # ref is the path to the reference database to use for taxonomy assignment
  ref: ""
  # taxfile is the path to the taxonomy file that corresponds to the reference database
  taxfile: ""
  ranks: []
# The epa-ng parameter specifies the parameters to use for the epa-ng taxonomy
# assignment tool.
epa-ng:
  tree: ""
  tree_format: "newick"
  msa: ""
  msa_format: "fasta"
  model: "GTR+G+F"
  heuristic: "dyn-heur"
  ref_taxonomy: ""
  chunk_size: 5000
  tree_ranks: 
    - "Kingdom"
    - "Phylum"
    - "Class"
    - "Order"
    - "Family"
    - "Genus"
    - "Species"
  gappa:
    distribution_ratio: -1
    consensus_thresh: 1

# Below are settings that control how ASVs that are unclassified by SINTAX will
# be handled. The default is to use the EPA-NG results to reassign the SINTAX
# taxonomy at high levels in the taxonomy tree.
reassign:
  # placeholder_rank is the taxonomic rank at which ASVs must be classified by EPA-NG, 
  # but unclassified by SINTAX in order to be reclassified.
  placeholder_rank: "Class"
  # placeholder_taxa is a list of taxa that ASVs must be classified as by both EPA-NG,
  # and SINTAX (at placeholder_rank) in order to be reassigned at lower levels.
  placeholder_taxa: ["Collembola", "Diplura", "Protura", "Insecta"]
  # reassign_rank is the taxonomic rank at which ASVs will be reassigned if they meet
  # the criteria above.
  reassign_ranks: ["Order"]
  # downstream_ranks is a list of ranks which will be updated for ASVs with reassignments
  downstream_ranks: ["Family", "Genus", "Species", "BOLD_bin"]


# The evaluation rank parameter specifies what taxonomic rank to use for evaluating
# the clustering results. This rank should correspond to the taxonomic rank that
# you expect the clustering to represent, e.g. if you expect the clustering to
# represent species, set this parameter to "Species".
evaluation_rank: "Species"
# Ranks to use for calculating consensus taxonomy for clusters
consensus_ranks:
  - "Family"
  - "Genus"
  - "Species"
# % threshold to use for calculating consensus taxonomy for clusters
# As an example, if a cluster has 10 sequences, and at least 8 of them are classified
# as "Species X" at the Species level, the consensus taxonomy for the cluster
# will be "Species X" if the threshold is set to 80, otherwise taxonomy will
# be attempted to be resolved at Genus level etc.
consensus_threshold: 80

########################################
##### CHIMERA FILTERING PARAMETERS #####
########################################
chimera:
  # Should chimera detection and removal be performed on the input data?
  remove_chimeras: True

  # Set a run name for the chimera detection and removal step
  run_name: "chimera1"
  
  # Select method for chimera filtering:
  #'batchwise' = run chimera detection on dataset as a whole.
  #'samplewise' = split input fasta into one file per sample and run chimera
  #               detection on each sample individually.
  method: "samplewise"
  
  # Select algorithm to use, you can choose from 'uchime_denovo', 'uchime2_denovo'
  # and 'uchime3_denovo'.
  algorithm: "uchime_denovo"
  
  # In batchwise method, require that a sequence marked as chimeric is present
  # with its parents in at least <min_samples_shared> samples
  min_samples_shared: 1
  
  # In batchwise method, require that a sequence marked as chimeric is present
  # with its parents in at least <min_frac_samples_shared> fraction of samples
  min_frac_samples_shared: 0
  
  # In samplewise method, require that a sequence is marked as chimeric in at least
  # <min_chimeric_samples> in order for it to be removed from analysis. If this
  # value is set to 0, ASVs have to be marked as chimeric in all samples
  min_chimeric_samples: 0
  
  # In samplewise method, require that a sequence is marked as chimeric in at least
  # <min_frac_chimeric_samples> fraction of samples in which the ASV is present in 
  # order for it to be removed from analysis. If this value is set to 0, ASVs have 
  # to be marked as chimeric in all samples
  min_frac_chimeric_samples: 0
  
  # Thec parameters below are specific to how the chimeric score of sequences is
  # calculated. Please see the Uchime
  # [docs](https://www.drive5.com/usearch/manual6/UCHIME_score.html) for
  # details. Note that these are not used for uchime2_denovo or uchime3_denovo.
  # Instead these algorithms require 'perfect chimeras' (see the Uchime manual
  # for more info)
  dn: 1.4
  mindiffs: 3
  mindiv: 0.8
  minh: 0.28
  #abskew: 2.0

###################################
##### CLUSTERING TOOLS TO RUN #####
###################################
# Specify what clustering software to use. Choose from 'swarm', 'opticlust',
# 'dbotu3'.
software:
  - "swarm"
  #- "opticlust"
  #- "dbotu3"

####################################
##### TOOL-SPECIFIC PARAMETERS #####
####################################
vsearch:  
  # minimum pairwise identity to report
  id: 0.84
  
  # pairwise identity definition (see vsearch manual for details and choices)
  iddef: "1"

  # threshold for query coverage
  # reject hits if fraction of query aligned is less than this value
  query_cov: 0.9

opticlust:
  # For opticlust, choose whether pairwise alignments should be generated with
  # 'vsearch' or 'mothur'
  aligner: "vsearch"
  
  # The delta parameter sets the stable value for the metric in opticlust
  delta: 0.0001
  
  # at what similarity threshold clusters are generated. 
  cutoff: 0.05

  # The `initialize` parameter sets the initial randomization for opticlust.
  # The default is `singleton` where each sequence is randomly assigned to its
  # own OTU. The other accepted setting `oneotu` means that all sequences are
  # assigned to one otu. 
  initialize: "singleton"

  # the floating point precision for opticlust.
  precision: 1000

swarm:
  # number of differences allowed to cluster sequences
  differences: 15

  # setting 'no-otu-breaking' to True deactivates the built-in cluster
  # refinement when d=1
  no-otu-breaking: False

  # if 'fastidious' is set to True swarm performs a second clustering pass to
  # reduce the number of small clusters when d=1
  fastidious: False

  # when 'fastidious' is set to True 'boundary' defines the minimum abundance of
  # what should be considered a large cluster.
  boundary: 0

  # the settings below modify the pairwise global alignment scoring parameters
  # when d>1
  match-reward: 5
  mismatch-penalty: 4
  gap-opening-penalty: 12
  gap-extension-penalty: 4

dbotu3:
  # dist sets the maximum allowed genetic dissimilarity between sequences
  dist: 0.1

  # abund sets the minimum fold difference for comparing two OTUs
  abund: 10.0

  # pval sets the minimum p-value for merging ASVs
  pval: 0.0005

###########################
##### NOISE FILTERING #####
###########################
noise_filtering:
  # run_name sets the name of the run for the noise filtering step
  run_name: "noise1"

  # split_rank sets the taxonomic rank to split the ASVs by before running the
  # noise filtering.
  split_rank: "Order"

  # assignment_rank sets the taxonomic rank to use for the noise filtering.
  # ASVs unassigned or ambiguously assigned at this rank are considered noise and will be removed.
  assignment_rank: "Order"

  # maximum number of target sequences to return for aligner
  max_target_seqs: 10

  # min_match is the minimum threshold of sequence similarity for considering any
  # OTU as an error of another
  min_match: 84

  # n_closest is the number of potential parent clusters to use in the comparison
  # when identifying noise
  n_closest: 10

  # echo_min_overlap (numeric on the interval (0,1)): In the echo filter, the minimum fraction of
  # the putative “noise” OTU samples that must also contain the putative
  # authentic CO1 “parent” OTU
  echo_min_overlap: 0.9

  # echo_read_ratio_type (“max”/“mean”): In the echo filter, the type of read
  # ratio considered by the echo_max_read_ratio criterion: “max” = maximum ratio
  # between noise and parent reads across the samples where they co-occur;
  # “mean” = mean ratio between noise and parent reads across the sample where
  # they co-occur.
  echo_read_ratio_type: mean

  # echo_max_read_ratio (numeric, positive real): In the echo filter, the
  # maximum value of the ratio between noise and parent reads
  echo_max_read_ratio: 0.1

  # echo_require_corr (TRUE/FALSE): In the echo filter, should a significant
  # correlation between noise and parent read numbers be required before the
  # putative OTU is removed? The significance is evaluated using a linear model,
  # requiring that the regression coefficient is smaller than
  # echo_max_read_ratio, and the p-value <=0.05.
  echo_require_corr: False

  # evo_local_min_overlap (numeric on the interval (0.1)): In the local evo
  # filter, minimum fraction of the putative “noise” OTU samples that must also
  # contain the putative authentic CO1 “parent” OTU.
  evo_local_min_overlap: 0.95

  # dist_type_local (“dadn”/“wdadn”): In the local evo filter, should distances
  # between representative ASVs be based on unweighted amino acid distances
  # (“dadn”) or amino acid distances weighted based on biochemical similarities
  # (“wdadn”)?
  dist_type_local: dadn

  # dist_threshold_local (numeric, positive real): In the local evo filter, the
  # distance value above which an OTU is considered to represent noise.
  dist_threshold_local: 1.8

  # dist_threshold_global (numeric, positive real): In the global evo filter,
  # the distance value above which an OTU is considered to represent noise.
  dist_threshold_global: 4.0

  # abundance_cutoff_type (“sum”/“max”): Should the abundance cutoff be based on
  # the sum of reads across samples in the dataset (“sum”), or the max number of
  # reads (“max”)?
  abundance_cutoff_type: sum

  # abundance_cutoff (numeric, positive integer): The minimum number of reads
  # required for an OTU to pass the abundance filter
  abundance_cutoff: 3
  
  # codon table to use for translation when aligning sequences with pal2nal
  codon_table: 5
  
  # reading frame start (1 based)
  codon_start: 2