FastG format?

[kbradnam]

FASTG v1.0 spec is now available from here:

http://fastg.sourceforge.net

[vmakinen]

I am not saying mapping to FastG is not possible, I am asserting that FM-indexes are not suitable (yet) for multiple hapolotypes in the same reference sequence.

Actually they are already suitable. A slight modification to BWT is enough:

http://www.cs.helsinki.fi/group/suds/gcsa/

[adaptivegenome]

The reference genome will still be relevant even if we could get a very good assembly. After all, the annotations are in the reference coordinate. To annotate a new assembly, we need to map the assembly to the reference genome.

We all wish to map data to a graph, but few have a clear definition of the problem, let alone the solution. Adopting graph alignment is likely to take longer than we hope. For now, my vague vision is a graph alone is not enough. We also need the alignment between the graph and the reference.

As to fastg, you can read from the archive that I a little worry about its scope (final scaffold only or generic sequence graph?), technical complexity (simpler and easier to parse format?) and mathematical clarity (more straightforward graph interpretation?), but probably it is me who has the wrong opinions.

I would be the first to admit that I am probably underestimating the complexity here, but a graph approach would be really nice.

I suppose the final specs are not released yet, however from the conference it seems that the format is very easy to parse and represents an obvious advance from an IUPAC coded reference (you can explicitly define indels, repeats, etc.).

[lh3]

The reference genome will still be relevant even if we could get a very good assembly. After all, the annotations are in the reference coordinate. To annotate a new assembly, we need to map the assembly to the reference genome.

We all wish to map data to a graph, but few have a clear definition of the problem, let alone the solution. Adopting graph alignment is likely to take longer than we hope. For now, my vague vision is a graph alone is not enough. We also need the alignment between the graph and the reference.

As to fastg, you can read from the archive that I a little worry about its scope (final scaffold only or generic sequence graph?), technical complexity (simpler and easier to parse format?) and mathematical clarity (more straightforward graph interpretation?), but probably it is me who has the wrong opinions.

[nilshomer]

You would know better than me how fast the technology is progressing, so I can't say for sure that it would be worth it, but I think FastG could be useful in specifying alternate reference sequences during mapping. I am not sure it would require significant alteration to existing methods.

I am not saying mapping to FastG is not possible, I am asserting that FM-indexes are not suitable (yet) for multiple hapolotypes in the same reference sequence.

[adaptivegenome]

It will be quite difficult to adapt the FM-index (BWT) based aligners. My prediction would be full-on assembly being the norm in about 2 years.

You would know better than me how fast the technology is progressing, so I can't say for sure that it would be worth it, but I think FastG could be useful in specifying alternate reference sequences during mapping. I am not sure it would require significant alteration to existing methods.

[nilshomer]

Thanks, this is great. I think the format might be also useful in mapping. However I do realize at some point we won't be mapping to a reference anymore.

It will be quite difficult to adapt the FM-index (BWT) based aligners. My prediction would be full-on assembly being the norm in about 2 years.

[adaptivegenome]

Thanks, this is great. I think the format might be also useful in mapping. However I do realize at some point we won't be mapping to a reference anymore.

[maubp]

It is being discussed for the next Assemblathon competition, and the mailing list was recently made public. See:

http://assemblathon.org/pages/mailing-list

https://lists.ucdavis.edu/sympa/arc/assemblathon-file-format

[adaptivegenome]

I saw some posters at Biology of Genomes this year mentioning the new FastG format for assemblers. I was wondering if anyone has heard about this and if a spec was available yet.