bin3C - simulated community and associated sequencing datasets

We simulated a human gut microbiome comprising 63 genomes from the GTDB annotated with an isolation source of faeces. No two genomes are more than 96% similar in terms of ANI.

A Generalized Pareto distribution was used to model an abundance profile, which was assigned in random order to the references. There is a 50:1 difference between the most and least abundant member.

Illumina shotgun and Hi-C reads were simulated using MetaART and sim3C (https://github.com/cerebis/sim3C).

A sweep was performed over depth of coverage, by serially subsampling initial high depth readsets. Shotgun depth was parameterised by the most abundant at 250x, while Hi-C was parameterised by the number of pairs (200 million pairs).

Shotgun was subsampled once, at half depth (125x), while Hi-C was subsampled 4 times (12.5, 25, 50, 100, 200 million pairs).

The random seed used throughout was 12345.

These simulated readsets were then analyzed using bin3C to retrieve metagenome-assembled genomes (MAGs). The resulting genome bins were validated using CheckM to estimate completeness and contamination.