Better handling of asymmetric library distributions

The distribution of jumping library fragment (“insert”) sizes can be far from Gaussian. Despite this, assemblers often model these distributions as Gaussian, leading to errors and inaccuracies. As of revision 37618,  ALLPATHS-LG calculates the actual pair separation distribution for all libraries by alignment to an intermediate assembly, and then uses these empirical distributions to estimate gaps between scaffolds. For libraries that are far from Gaussian this results in much more accurate gap estimates, and this in turn allows for better gap patching, leading to longer contigs.

This entry was posted in Release. Bookmark the permalink.

Comments are closed.