Following the strategy described above for correcting relocation errors, if the distance between two contigs in the predicted assembly was less than MIL in the reference genome, this relocation was correctable.
*: if the interval between two contigs is smaller than MIL in the reference genome but they are joined together in the predicted assembly, this relocation is considered correct.
Second, if there still is more than one match between two contigs, only the longest is retained.
Read-pairs were tested separately to check if they could constrain the scaffolding problem by defining a unique path between two contigs.
Unreliable linkages between two contigs were filtered and the remaining contigs with compatible connections to each other, and having at least three read-pairs, were constructed into scaffolds.
Vertices represent Newbler contigs and edges link between two contigs.
*gap between two contigs not taken into account.
In order to generate scaffolds, contigs were connected using "N" to represent unknown sequences between two contigs.
All ESTs from six Butte 86 contigs (#3, 5, 6, 7, 9,11) were found in single contigs in HarvEST 1.14, while ESTs from four contigs (#1, 2, 4, 10) were divided between two contigs and ESTs from contig #8 were distributed among three different contigs.
Otherwise, 100 Ns are inserted between two contigs just to indicate a possible gap.
