- 2016/6/17 first update, next and previous alignment name were moved to mandatory fields
- 2016/6/11 proposal of a GPA format, Hajime Suzuki
Graphical pairwise alignment (GPA) format is a text-based line-separated structure to represent a set of pairwise local alignments between two graphically-represented sequences. Although the SAM and BAM format are widely used as de-facto standard output formats of the sequence alignment tools, the SAM and BAM format has a defect in lacking a path information on the read side, making it unable to represent graphically-represented query sequences without external files or complecated optional fields. The GPA format is designed to remove the defect and complexity of the SAM format, following the convention of the GFA format, representing each record in a line and each field in a record delimited with tabs ('\t'). The format is originally used as a default output format of the comb aligner and libaw alignment writer library.
The entire file consists of a single header line and multiple alignment lines. The field representation of the GPA format largely follows that of the Graphical Fragment Assembly (GFA) format, that is, each line starts with a single-character identifier, the fields are separated by tabs, and optional fields are distinguished by two-letter identifiers followed by a type character with a colon (:) between them.
- A single printable ASCII character
- i signed integer
- Z ASCII string (spaces allowed)
The other types found in the original GFA format, f, H, and B are reserved but not used in the current version.
Each record has a optional fields at the tail of the line, separated with tabs. Each optional filed consists of a tuple of a tag, a type, and a value. The tag is a two-letter identifier followed by a type identifier character and content of the field delimited with colons.
The header line is identified with a H character. The line have no mandatory field other than the line identifier and an optional version number field (VN).
- identifier:A must be
'H'
- VN:Z string of the program version number
(is SO:Z (sorting order) required?)
H VN:Z:0.1
The alignment line is identified with a 'A' character, having 10 mandatory fields to represent an alignment path and its projection (positions and lengths) to the two sequence segments.
- identifier:A must be
'A' - name:Z unique name of the alignment (string w/spaces)
- aname:Z name of the sequence segment A
- apos:i segment-local index (0-origin) of the alignment start position on the sequence segment A
- alen:i length of the alignment on the sequence segment A
- adir:A direction of the alignment,
'+'(forward) or'-'(reverse) - bname:Z name of the sequence segment B
- bpos:i segment-local index of the alignment start position on the sequence segment B
- blen:i length of the alignment on the sequence segment B
- bdir:A direction of the alignment on the sequence segment B
- cigar:Z CIGAR string of the alignment
- prev name of the previous (upstream) alignment
- next name of the next (downstream) alignment
- MQ:i mapping quality
A 0 sec0 0 4 + sec1 0 4 + 4M * * MQ:i:255
The example below represents a set of mappings of three sequences to a graphical reference structure with four sections. The read 1 aligns to the A->B->D path of the reference, 2 to the A->D, and 3 to the D->C->A (on the reverse-complemented side), respectively.
H VN:Z:0.1
A 0 A 0 4 + read1 0 4 + 4M * 1 MQ:i:255
A 1 B 0 4 + read1 4 4 + 4M 0 2 MQ:i:255
A 2 D 0 8 + read1 8 8 + 8M 1 * MQ:i:255
A 3 A 0 4 + read2 0 4 + 4M * 4 MQ:i:255
A 4 D 0 8 + read2 4 8 + 8M 3 * MQ:i:255
A 5 D 4 4 - read3 16 4 + 4M * 6 MQ:i:255
A 6 C 12 12 - read3 4 12 + 12M 5 7 MQ:i:255
A 7 A 4 4 - read3 0 4 + 4M 6 * MQ:i:255