Fig. 5

K-mer spectra of all samples. Given input reads and a set of contigs assembled from the reads, let \(N^{(c)}_x\) be the number of k-mers occurring \(\ge x\) times in reads and exactly c times in contigs (thus “right-accumulated”). Then, \(N_x=\sum _c N^{(c)}_x\) is the number of k-mers occurring \(\ge x\) times in reads. Plots on the left column show the k-mer spectra of \(\ge\) 1 Mb circular contigs. The height (i.e., the length on the Y-axis) of the blue area intersecting at x equals \(N^{(0)}_x/N_x\). It is the fraction of read k-mers occuring \(\ge x\) times in reads but absent from the assembly. The height of the orange area at x equals \(N^{(1)}_x/N_x\). The white dashed line shows \(N_x\) and each plot is truncated at \(x_t\) where \(N_{x_t}<10^6\). Intuitively speaking, a large blue region in the right part of a plot suggests an incomplete assembly that misses many high-abundance k-mers in reads. Plots on the right column show the k-mer spectra of hmBin MAGs. The light orange area with forward hatches indicates the contribution of rescued circles and the brown area with backward hatches indicates the contribution of MAGs found by non-topology based binning. The blue area is generally smaller due to the inclusion of non-circular MAGs