cschin

Some thought about Heng Li's proposal for assembly graph format http://lh3.github.io/2014/07/19/a-proposal-of-the-grapical-fragment-assembly-format/


some quick comments.

Is this format trying represent the raw overlaps or finally assembly graph or both?

It seems to me that it is more suitable for the first. In the work to represent diploid genome assembly, I had to do multiple level of reduction of the graph from the initial string/overlap graph to simply the problem. if we are looking at a more reduced assembly, we might have to deal with edges corresponding to unitigs with the same in and out nodes. In this format, such bubble paths (difference between them bigger than small indel) will be in different row, the behavior of such edges with the same in and out node should be defined. What I did for diploid work is to assign uid for each edges.

Also, I do think the final assembly should avoid the bidirectional edges. It should be resolved by the assembler. From pragmatic point, it will confuse a lot of bi

cschin

(1) If we use Celera Assembler's bogart unitiger, it will generate a file called "bests.edges" in the "4-unitigger" as one of its output directories.

(2) I wrote a simple script converting the edge list to a GML. The script can be downloaded https://github.com/PacificBiosciences/HBAR-DTK/blob/master/src/CA_best_edge_to_GML.py  

(3) Load the graph into gephi (https://gephi.org/)

(4) I typically use the following steps of different layout algorithms in Gephi to get a good layout, 

1) "YifanHu's Multilevel" to get a rough layout, the output usually catches good large scale structure and detangle the graph reasonably one so I can start to see features of the assembly overlap graph (or the string graph.)

cschin