Meaning of gfa outputs - Githubissues

PacificBiosciences / FALCON

FALCON: experimental PacBio diploid assembler -- Out-of-date -- Please use a binary release: https://github.com/PacificBiosciences/FALCON_unzip/wiki/Binaries

https://github.com/PacificBiosciences/FALCON_unzip/wiki/Binaries

Other

204 stars 103 forks source link

Meaning of gfa outputs #682

Closed gbdias closed 5 years ago

gbdias commented 5 years ago

Hi,

In the latest version of Falcon (falcon-unzip 1.1.4, falcon-kit 1.2.4, pypeflow 2.1.1), the assembly output include multiple files with the gfa extension.
```
asm.gfa
asm.gfa2
contig.gfa2
sg.gfa
sg.gfa2
```
Could you provide an explanation on what are these files encoding or links to the relevant documentation?
I did some extensive searching in the Falcon docs but could not find any info about the gfa files.
I also tried examining these files using Bandage. The only file with a number of nodes compatible with the number of contigs contained in my assembly (contig.gfa2) had zero edges, so I am not sure how this compares to Canu's .contigs.gfa, where you can see chromosome-like graphs comprising multiple connected nodes.

Thanks, Guilherme

foala commented 5 years ago

Hi, asm = assembly pread graph sg = string graph contig = contig graph with associate contigs attached to each primary contig

GFA is the standard format used for graph representation. GFA2 is the new format. Both are supported by Bandage.

Source: https://github.com/PacificBiosciences/FALCON/blob/master/falcon_kit/bash.py

Hope this helps

gbdias commented 5 years ago

Hi @pb-cdunn,

I am still unsure on how to visualize the FALCON gfa files.

None of the gfa files generated by FALCON have a total length that matches the assembly length when loaded onto Bandage.
- Assembly length: 170,967,379 (p_ctgs) + 45,710,955 (a_ctgs).
- sg.gfa: 1,089,627,645
- asm.gfa: 893,130,356
- contigs.gfa: 10,410 bp
Also, the file where the node count reflects the number of contigs (contigs.gfa) has zero edges or edge overlaps. And the node lengths don't reflect contig lengths.
Is there a similar view to that of Canu contigs.gfa available from FALCON output? Like in the figure below from Koren et al. 2017. Gen Res