While the genome-aligned reads from the initial LY2109761 TGF-beta inhibitor library NVP-BKM120 collapsed to 25.7% of the original dataset, the genome-aligned reads from the round 3 sub-library collapsed less, to 38.2% of the original dataset. These results indicate that the long march reduced the redundancy of the initial cDNA library. In addition to contig size, the advantage to total genome coverage provided by the long march was examined. Several datasets of randomly sampled genome-aligned reads from the round 3 sub-library and from the initial library were mapped back to the P. falciparum genome and the number of genomic bases covered by at least one read was measured for each dataset. Even with a small dataset of 50,000 reads, the round 3 sub-library covered 35% more genomic bases than the initial library. As the number of reads in each dataset grew, so too did the difference in coverage. At 500,000 reads apiece, the marched sub-library vastly outpaced the initial library by covering an additional 1.1 million bases, an increase in coverage of 39%. The long march protocol was also applied to RNA extracted from a serum specimen from a patient with HBV-related acute liver failure in order to assess its applicability to metagenomic analysis. 36 bp reads from the initial library as well as the round 3 sub-library were aligned to the HBV genome using ELAND. Sequencing of the round 3 sub-library generated a greater percentage of location-collapsed HBV reads than were generated by sequencing the corresponding initial library. This trend translated to enhanced genome coverage of HBV�Cwith a dataset of 300 genome-aligned reads, the round 3 sub-library covered 42% more genomic bases than the initial library. Thus the long march increases coverage of a target genome in both resequencing and metagenomic contexts. We used theoretical considerations to assess the utility of the long march protocol for de novo genome or metagenome assembly as well. For such assembly to be reliable, the length of overlap between any two reads must be sufficient to identify their common origin. In the initial P. falciparum library, the extent of overlap between reads decayed exponentially and therefore included many instances of both insufficient overlap for de novo assembly and excess overlap for minimal contig extension. In the long march procedure, a step size can be selected that creates the minimum overlap between adjacent steps necessary for correct assembly given the read length and dataset size. To avoid spurious joining, datasets with many unique sequences required longer overlaps than those with few unique sequences. Modeling and simulation of the assembly process revealed amplicon library complexity to be critical to the assembly of marched reads into contigs. The benefit gained from optimization of overlap length requires the sequencing of all steps from a given library amplicon within a reasonable number of reads. With increasing complexity of the template pool, this stipulation becomes less likely.