RadicalSAM.org: A Resource to Interpret Sequence-Function Space and Discover New Radical SAM Enzyme Chemistry

Abstract

The radical SAM superfamily (RSS), arguably the most functionally diverse enzyme superfamily, is also one of the largest with ~700K members currently in the UniProt database. The vast majority of the members have uncharacterized enzymatic activities and metabolic functions. In this Perspective, we describe RadicalSAM.org, a new web-based resource that enables a user-friendly genomic enzymology strategy to explore sequence-function space in the RSS. The resource attempts to enable identification of isofunctional groups of radical SAM enzymes using sequence similarity networks (SSNs) and the genome context of the bacterial, archaeal, and fungal members provided by genome neighborhood diagrams (GNDs). Enzymatic activities and in vivo functions frequently can be inferred from genome context given the tendency for genes of related function to be clustered. We invite the scientific community to use RadicalSAM.org to (i) guide their experimental studies to discover new enzymatic activities and metabolic functions, (ii) contribute experimentally verified annotations to RadicalSAM.org to enhance the ability to predict novel activities and functions, and (iii) provide suggestions for improving this resource.

Keywords: Radical SAM superfamily; functional assignment; genome neighborhood diagrams; genomic enzymology; isofunctional families; sequence similarity networks; web resource.

Figure 1

Despite high diversity in sequence and reaction outcome, all members of the RSS generate 5′-deoxyadenosyl (5′-dA) radical using a conserved [4Fe–4S]-forming motif that binds and reductively liberates Met from SAM.⁻ RadicalSAM.org provides easy access to a genomic enzymology strategy to catalog known enzymatic activities and discover new ones within the RSS.

Figure 2

SSN generated with a maximum e-value edge threshold of 1e-20 used by the SFLD to identify its 20 functionally characterized subgroups (colored/numbered clusters) and 22 uncharacterized subgroups. Large nodes represent experimentally characterized proteins: downward arrows indicate a structurally characterized protein; diamonds indicate no structural characterization. Reproduced with permission from ref (9). Copyright 2018 Elsevier.

Figure 3

Home page of RadicalSAM.org.

Figure 4

SSN of the starting point for RSS subgroup identification. The network is visualized with 11 as the minimum edge alignment score threshold and colored based on previously assigned SFLD subgroups (Figure 2 and Table 1).

Figure 5

Five megaclusters (containing multiple SFLD subgroups) and five standard clusters (containing single SFLD subgroups) after manual deletion of “long” edges that correspond to larger alignment scores and connect functionally divergent nodes/subgroups in the large cluster in Figure 4. The nodes are colored based on previously assigned SFLD subgroups (Figure 2 and Table 1).

Figure 6

Segregated subgroups in RadicalSAM.org. The nodes are colored based on previously assigned SFLD subgroups (Figure 2 and Table 1). (A) UniRef50 SSNs for Megaclusters-1-2 through -1-8 were removed from Megacluster-1 using an alignment score edge threshold of 30 and displayed with 16 as the minimum edge alignment score threshold. The remaining Megacluster-1-1 is displayed using 11 as the minimum edge alignment score threshold. (B) UniRef50 SSNs for Megaclusters-2-1 through -2-7 were segregated using 12 as the minimum edge alignment score threshold. (C) UniRef90 SSNs for Megaclusters-3-1 through -3-12 segregated using 18 as the minimum edge alignment score threshold. (D) UniRef90 SSNs for Megaclusters-4-1 through -4-11 segregated using 22 as the minimum edge alignment score threshold. (E) UniRef50 SSNs for Megaclusters-5-1 through -5-3 were segregated using 12 as the minimum edge alignment score threshold. (F) UniRef50 SSNs for Clusters 6-10 from the segregated SSN in Figure 5.

Figure 7

Example of an Explore page using Megacluster-3-1, the 7-carboxy-7-deazaguanine synthase-like radical SAM proteins (SFLD subgroup 1).

Figure 8

Example of a taxonomy sunburst display. Shown is Megacluster-3-1, 7-carboxy-7-deazaguanine synthase-like (SFLD subgroup 1). Green, bacteria; orange, archaea; magenta, eukaryota.

Figure 9

Representative GND Viewer page: Cluster-6 in the “diced” SSN for Megacluster-1-1 generated with a minimum edge alignment score threshold of 60.

Figure 10

Portion of the Skylign display of the HMM for Megacluster-3-1, 7-carboxy-7-deazaguanine synthase-like, SFLD subgroup 1, showing the conserved CX₃CX₂C motif that binds SAM.

Figure 11

Representative diced SSNs for Megacluster-1-1 (SPASM/twitch domain-containing).

Figure 12

Navigation through diced megaclusters. (A) “Previous” and “Next” navigation buttons (green arrows) direct the cluster selection backward and forward, respectively, in the selected diced SSN. (B) “Cluster” drop down menu allows the user to select any cluster in the currently viewed diced SSN. (C) Alignment score drop down menu allows the user to view the cluster in the selected diced SSN.

Figure 13

AS Walk-Through pop-up window showing the identity of the progenitor cluster (“Previous Cluster) and progeny clusters (“Next Clusters”).

Figure 14

Search functions: “Find by UniProt ID” identifies the cluster(s) containing the user-specified accession ID; “Find by Sequence” identifies the cluster(s) with the best HMM match to the user-specified sequence; “GND Lookup” provides the GND(s) for the user-specified UniProt ID(s); “Find by Taxonomy” provides a list of the UniProt accession IDs and their clusters for the user-specified genus/species.

Figure 15

Submit page for community submission of enzymatic activities and metabolic functions.

Figure 16

Contact page form for submitting feedback or questions.