Fig. 1.

AmpliCI: inferring ASVs from samples. (1) Construct unique sequence set $\mathcal{S}$, and put the most abundant unique sequence in haplotype set ${\mathcal{H}}_1$. (2) Given the current haplotypes ${\mathcal{H}}_k$, the scaled abundances ${\tilde{a}}_{{\mathit{s}}_m}^{(k)}$ are estimated for each remaining unique sequence ${\mathit{s}}_m$ via update function (5), and the haplotype candidate ${\mathit{s}}_m$ with highest scaled abundance is selected. (3) Verify the approximate BIC improves and the diagnostic probability p_mz is small enough, and update ${\mathcal{H}}_{k+1}=\{{\mathcal{H}}_k,{\mathit{s}}_m\}$. Otherwise, permanently discard candidate ${\mathit{s}}_m$ and select the next most abundant candidate. (4) Repeat (2)–(3) until the scaled abundance ${\tilde{a}}_{{\mathit{s}}_m}^{(K)}$ of all remaining unique sequences ${\mathit{s}}_m$ are below the user-determined abundance threshold c. (5) Output the K haplotypes ${\mathcal{H}}_K$