SCAPP: an algorithm for improved plasmid assembly in metagenomes

Abstract

Background: Metagenomic sequencing has led to the identification and assembly of many new bacterial genome sequences. These bacteria often contain plasmids: usually small, circular double-stranded DNA molecules that may transfer across bacterial species and confer antibiotic resistance. These plasmids are generally less studied and understood than their bacterial hosts. Part of the reason for this is insufficient computational tools enabling the analysis of plasmids in metagenomic samples.

Results: We developed SCAPP (Sequence Contents-Aware Plasmid Peeler)-an algorithm and tool to assemble plasmid sequences from metagenomic sequencing. SCAPP builds on some key ideas from the Recycler algorithm while improving plasmid assemblies by integrating biological knowledge about plasmids. We compared the performance of SCAPP to Recycler and metaplasmidSPAdes on simulated metagenomes, real human gut microbiome samples, and a human gut plasmidome dataset that we generated. We also created plasmidome and metagenome data from the same cow rumen sample and used the parallel sequencing data to create a novel assessment procedure. Overall, SCAPP outperformed Recycler and metaplasmidSPAdes across this wide range of datasets.

Conclusions: SCAPP is an easy to use Python package that enables the assembly of full plasmid sequences from metagenomic samples. It outperformed existing metagenomic plasmid assemblers in most cases and assembled novel and clinically relevant plasmids in samples we generated such as a human gut plasmidome. SCAPP is open-source software available from: https://github.com/Shamir-Lab/SCAPP . Video abstract.

Keywords: Assembly; Plasmids.

Fig. 1

Graphical overview of the SCAPP algorithm. (A) The metagenomic assembly graph is created from the sample reads. (B) The assembly graph is annotated with read mappings, presence of plasmid specific genes, and node weights based on sequence length, coverage, and plasmid classifier score. (C) Potential plasmids are iteratively peeled from the assembly graph. An efficient algorithm finds cyclic paths in the annotated assembly graph that have low weight and high chance of being plasmids. Cycles with uniform coverage are peeled. (D) Confident plasmid predictions are retained using plasmid sequence classification and plasmid-specific genes to remove likely false positive potential plasmids

Fig. 2

Evaluating and peeling cycles. Numbers inside nodes indicate coverage. All nodes in the example have equal length. A Cycles (a,e,f) and (c,e,g) have the same average coverage (13.33) and coefficient of variation (CV, 0.35), but their discounted CV values differ: The discounted coverage of node a is 6, and the discounted coverage of node e is 10 in both cycles. The left cycle has discounted CV=0.22 and the right has discounted CV=0. By peeling off the mean discounted coverage of the right cycle (10) one gets the graph in B. Note that nodes g,c were removed from the graph since their coverage was reduced to 0, and the coverage of node e was reduced to 10

Fig. 3

Annotation of genes on the plasmids identified by SCAPP in the human gut plasmidome sample. A Functional annotations of the plasmid genes. B Host annotations of the plasmid genes. “Broad-range” plasmids had genes annotated with hosts from more than one phylum

Fig. 4

Outline of the read-based performance assessment. Plasmidome (I) and metagenome reads (II) are obtained from subsamples of the same sample. (III) The metagenome reads are assembled into a graph. (IV) The graph is used to detect and report plasmids by the algorithm of choice. (V) The plasmidome reads are matched to assembled plasmids. Matched plasmids (red) are used to calculate plasmid read-based precision. (VI) The plasmidome reads are matched to the assembly graph contigs. Covered contigs (red) are considered plasmidic. The fraction of total length of plasmidic contigs included in the detected plasmids gives the plasmidome read-based recall

Fig. 5

Performance on the parallel datasets. A Plasmidome read-based performance. B Performance of each tool on the plasmids assembled from the metagenome using as gold standard the plasmids assembled from the plasmidome by the same tool. C Overall performance on the plasmids assembled from the metagenome compared to the union of all plasmids assembled by all tools in the plasmidome