|Quality, from class I to III chunks|
|% in class IV to V chunks|
|At 100-kb distance|
For five genomes, statistics from ALLPATHS, Velvet, and EULER assemblies are shown. Base accuracy: the inferred quality score for the bases in chunk classes I to III, obtained by dividing the number of errors in these chunks by the total number of bases, and then taking -10*log10 of this quantity. Misassemblies: total fraction of bases in chunks in classes IV to V. Long-range validity: we chose 10,000 pairs of 100-base sequences at random from the scaffolds, where the two sequences were separated by 100,000 bases, considering only cases where both sequences could be placed uniquely on the reference. We scored the placement as 'valid' if both sequences were placed on the same chromosome, in the same orientation, with separation 100,000 ± 25% bases. We report the fraction of valid placements. N/A: all scaffolds were of length <100 kb.