TABLE 1.

Overview and basic characteristics of AMR-MGE colocalization analysis using two general approaches: alignment and assembly.

Approach	Alignment-based			Assembly-based
Method	Ordinatation		Network analysis	De novo assembly	Resistome risk pipeline
Operation	NMDS	•Biplot Analysis	•Bayesian Network: Directed Acyclic Graph (DAG)	•Linear DNA Metagenomic Assembly	•Analysis Pipeline using IDBA-UA Assembly
Program	Vegan v2.5-5	•BNLearn v4.4.1	•MetaSPAdes	•MetaCompare
Input Data	•AMR/MGE accession count matrix filtered for sparse genes •Normalized (CSS) •Transformed (Hellinger)	•AMR/MGE accession counts, metadata variables (continuous or categorical) •Blacklisting non-sensical network relationships (arcs)	•Trimmed pair-end FASTQ files	•Trimmed pair-end FASTQ files
Database	Pipeline: AMRPlusPlus v 1.0.1 Plasmids: PlasmidFinder v 2.1 Prophage and other MGE : ACLAME v 0.4	AMR: MEGARes v 1.0 ICE: ICEBerg v 2.0	MEGARes v 1.0, PlasmidFinder v 2.1, ICEBerg v 2.0, ACLAME v 0.4	CARD, ACLAME v 0.4, PATRIC v 3.6.2
Analytical Parameters	•K (stress) < 0.2, good 2D representation in reduced dimensions •Anosim R > 0.75, well separated •Anosim R = 0.5, moderate overlap •Anosim R < 0.25, significant overlap	•M2 test statistic •P < 0.05	•Learning algorithm: Hill climb •BN Classifier: Naive Bayes •Conditional independence test: hybrid modeling •Arc blacklisting •Bootstraps: 1000 •Empirical threshold frequency: 0.70	•Default assembly parameters •BLAST to MGE or AMR databases for contigs >800 bp of a continuous stretch with >90% sequence identity and E-value < 1 × 10–10	•Default assembly parameters •BLAST alignment to MGE or AMR based on defaults
Colocalization	•2D spatial visualization of AMR/MGE genes (at any ontologic level) according to a metadata variable	•Overlay of AMR and MGE NMDS plots showing co-occurrence according to a metadata variable	•Predictive network of AMR and MGE gene co-occurrence and visualization of transitivity and centrality of critical genes/groups	•Mobility index (MI): Number of contigs containing an AMR gene type and MGE gene divided by total number of contigs containing an AMR gene type in a sample	•Risk score •3D Risk visualization
Method Ref.	•Ordination: (Oksanen, 2019) •CSS Normalization: (Paulson et al., 2013) •Hellinger Transformation: (Legendre and Gallagher, 2001)	•Bayes network analysis: (Scutari, 2010)	•Assembly defaults: (Nurk et al., 2017)	•Assembly defaults: (Peng et al., 2012) •Pipeline default: (Oh et al., 2018)

Basic characteristics of metagenomic colocalization are described.