In 454 sequencing, the project leader can save money by ordering many unpaired ("shotgun") libraries and few paired end libraries. Fewer mates should yield shorter scaffolds unless mate pair saturation has been reached. We can test whether this is the case. Re-run cgw a few times on the same data but each time remove some of the mate links. (Leave the reads, just remove their links.) Plot the scaffold N50 at several sample rates. Extrapolate past the 100% sample. That is, extend the curve beyond the current mate pair level. That will give a prediction of scaffold N50 growth given additional mate pair data.