de novo assembly (velvet or others)

[strob]

Dear all,

is velvet (or others) able to use a read twice in the assembly of two different contigs or does velvet need in this case two duplicates of this read?

It could be that in certain genes highly conserved domains are present. In order to be able to assemble the different genes, certain reads could be used more than once. Is this possible in velvet (or others)?

The reason why I ask this is because I filter my dataset for duplicated reads, thereby only keeping the "unique" ones. But I wonder now if I do not lose information by doing this. If velvet (or others) is able to use reads more than once, the filtering is fine; if not, I should keep all my reads.

Any comments more than welcome.

Many thanks in advance,

Steven

[jts]

Velvet and other de Bruijn graph assemblers (abyss, Euler-SR) are able to re-use reads. Each instance of the read contributes the same k-mer path to the graph so removing redundant copies does not effect the structure of the graph. That said, you should not filter the data as most assemblers track the number of times each k-mer exists in the input data. This information is used for error correction/removal and to infer the copy count for each segment in the graph which can help resolve repeats.