Tabular: split taxonomy string

Expand one semicolon-delimited lineage field into rank columns when the lineage tokens carry prefixes such as sk__, p__, and s__. The MAPseq-to-ampvis2 IWC workflow reads column 3, emits an OTU_ identifier from column 1, then writes superkingdom through species in a fixed order. Each input data row becomes one output row. A missing rank becomes an empty field, so downstream tools can rely on the same column positions.

Use prefix dispatch when ranks can be missing or appear in a different order. Splitting directly into positions is suitable only when the upstream format guarantees a fixed rank order and depth. A positional split cannot tell whether the second token is a kingdom or a phylum after one rank is omitted.

Input and output contract

The recipe below expects a tab-separated input with two non-data lines, an identifier in column 1, and the lineage in column 3. It drops both non-data lines. Its output has a header followed by nine tab-separated fields per data row: OTU, Superkingdom, Kingdom, Phylum, Class, Order, Family, Genus, and Species. It discards other input columns, including abundance. If abundance must survive, include its source field in the output header and print statement.

Connect the input dataset to infile, put the awk program in code, leave variables: [] when no values are injected, and publish outfile. The IWC step pins toolshed.g2.bx.psu.edu/repos/bgruening/text_processing/tp_awk_tool/9.5+galaxy0 at Tool Shed changeset 3dc70b59608c. Its header is added by a later awk step. This compact variant prints the header in the same step:

tool_id: toolshed.g2.bx.psu.edu/repos/bgruening/text_processing/tp_awk_tool/9.5+galaxy0
tool_state:
  infile: { __class__: ConnectedValue }
  variables: []
  code: |-
    BEGIN {
      FS = OFS = "\t"
      print "OTU", "Superkingdom", "Kingdom", "Phylum", "Class", "Order", "Family", "Genus", "Species"
    }
    NR > 2 {
      otu_id = "OTU_" $1
      superkingdom = kingdom = phylum = tax_class = order = family = genus = species = ""
      n = split($3, taxonomy, ";")
      for (i = 1; i <= n; i++) {
        token = taxonomy[i]
        if (token ~ /^sk__/) superkingdom = token
        else if (token ~ /^k__/) kingdom = token
        else if (token ~ /^p__/) phylum = token
        else if (token ~ /^c__/) tax_class = token
        else if (token ~ /^o__/) order = token
        else if (token ~ /^f__/) family = token
        else if (token ~ /^g__/) genus = token
        else if (token ~ /^s__/) species = token
      }
      print otu_id, superkingdom, kingdom, phylum, tax_class, order, family, genus, species
    }

The tool wrapper passes tab field separators to awk and writes outfile with the input’s datatype and metadata. The explicit FS = OFS = "\t" makes the program’s parsing and output separators clear. The IWC program instead joins its output fields with literal tab strings. The wrapper’s variables repeat is optional, and this recipe does not use it.

Adapt and check the result

• Headers: NR counts physical input lines. Change NR > 2 to NR > 1 for one header line, or remove the condition for headerless data. A wrong offset silently drops a data row or treats a header as taxonomy.
• Prefixes and blanks: The program retains tokens such as p__Firmicutes and leaves missing ranks empty. If the consumer expects bare names or a marker such as unassigned, transform those values explicitly and test the resulting contract. Prefixes must start at the beginning of each token. Trim whitespace first if the source inserts spaces after semicolons.
• Column shape: print uses OFS between all nine fields, including empty interior or trailing ranks. Check the first data row, the field count, and the expected datatype before connecting a downstream tool. A duplicate rank prefix in one lineage is overwritten by the last matching token.

The checked-in test named in verification_paths compares complete output bytes for reordered ranks and absent ranks. The pinned IWC workflow establishes the source idiom, while the local test checks this one-step variant. See tabular-prepend-header if header construction must remain a separate step, and tabular-cut-and-reorder-columns for later projection.