Abstract The consensus emerging from microbiome studies is that they are far more complex than previously thought, requiring deep sequencing. As deep sequenced datasets provide greater coverage than previous datasets, recovering a higher proportion of reads to the assembly is still a challenge. To tackle this issue, we set of to identify if multiple iterations of assembly would allow for otherwise lost contigs to be formed and studied and if so, how successful is such an avenue at improving the current methodology. A simulated metagenomic dataset was initially used to identify if multiple iterations of assembly produce useable contigs or mis-assembled artefacts were produced. Once we had confirmed that the secondary iterations were producing both accurate contigs without a reduction in contig quality we applied this methodology in the form of Spherical to 3 metagenomic studies. The additional contigs produced by Spherical increased the number of reads aligning to an identified gene by 11–109% compared to the initial iterations assembly. As the size of the dataset increased, as did the amount of data multiple iterations were able to add. Availability Spherical is implemented in Python 2.7 and available for use under a MIT licence agreement at: https://github.com/thh32/Spherical