Using ABySS Tools for Ad Hoc OLC Assembly

Question

[dbrowne.up]

Hello,

I have a set of contigs that I would like to re-assemble using an Overlap-Layout-Consensus (OLC)-type approach. Let's say the minimum contig size in the set is 1000 bp. If I understand correctly, the following commands should: 1) find all overlaps between 100-999 bp, and store that overlap information in the file INPUT-1.dot; 2) clean up the overlap graph, ignoring contigs in the file REPEATS.fa, writing the assembled unambiguous paths to the file INPUT-1.path and the new graph to INPUT-2.dot; then 3) merge unambiguously overlapping contigs using the information computed in the first two steps and write the merged sequences to INPUT-2.fa.

AdjList -v -k1000 -m100 INPUT-1.fa > INPUT-1.dot
abyss-filtergraph -v -k1000 --dot --no-SS --assemble -g INPUT-2.dot -i REPEATS.fa INPUT-1.dot INPUT-1.fa >INPUT-1.path
MergeContigs -v -k1000 -o INPUT-2.fa --merged INPUT-1.fa INPUT-2.dot INPUT-1.path

I know this is probably quite a crude approach, but would it achieve what I am trying to do? Any thoughts or suggestions otherwise are appreciated. I am happy to clarify any points that are unclear, if needed.

Thanks,
Dan

Answer 1

[benv]

Looks sensible to me. I didn't realize that abyss-filtergraph could be used for the layout stage of assembly -- that is kind of neat.

As an alternative to AdjList, you may wish to try abyss-overlap. And as an alternative to abyss-filtergraph you may wish to try abyss-layout. I'm not really sure what the trade-offs are between those pairs of programs -- just giving you a heads up that those options are available too.

[dbrowne.up]

Thanks Ben! Quick follow-up question: how are transitive edges and shims defined? Just wondering whether or not I should aim to filter those out of the overlap graph prior to layout and consensus.