Taxonomic Network

What Figure 4 Actually Shows

Figure 4: Core and specific microbiota for organic and conventional apples. Core microbiota (taxa occurring in 75% of all replicates) of each management group (conventional and organic) were combined for network analysis. To be included, taxa had to exhibit at least 0.01% abundance in the whole dataset. Node size correspond to relative abundance in the dataset as denoted in the legend on the bottom left, node labels display taxonomic identification of OTUs on genus level wherever possible and node color indicates appropriate phylum, as described in the legend on the top right. source: Wassermann et al. 2019

A clustering network based on the core taxa of the tissues of each apple management group was constructed to visualize the taxa present in all apples as well as the taxa that are specific for each management group (Figure 4). Only taxa occurring with at least 0.01% abundance in the whole dataset were included in the network analysis. All apples were found to share a high abundant core microbiota; 73 out of 141 genera were shared. Among them, highly abundant Proteobacteria were most dominant and abundant with 45 genera. In total 16 genera were found only in organically managed apples, and 50 genera, predominated by Proteobacteria (33 genera) were specific for conventional apples. source: Wassermann et al. 2019, Results text

Observation

Only taxa occurring in at least 3 of 4 biological replicates (75%), at genus level, are included in the figure, according to the caption.

Unknown

The figure caption states a 75%-of-replicates prevalence threshold, but the Results text describing the same figure only mentions the 0.01% abundance threshold and does not mention any prevalence criterion at all. This is a direct inconsistency between two descriptions of the same methodology within the paper itself, not just an ambiguity in how the threshold was applied. It leaves genuinely unclear what criteria actually produced the 73/16/50 genus counts reported in the text.

Unknown

Separately, even taking the caption's 75%-of-replicates threshold at face value, it is stated per management group but does not say whether this was applied within each individual tissue or pooled across all six tissues within a management group. These are meaningfully different criteria: a taxon present in most replicates of just one tissue could pass a per-tissue threshold while failing a pooled one, or vice versa.

Is This a Network Analysis?

A taxonomic co-occurrence network, in the usual microbiome sense, connects taxa to each other based on a measured statistical relationship, such as correlation or co-occurrence frequency, typically with edge weights and a stated method for how those relationships were calculated.

Observation

Figure 4 does not show any relationship between taxa themselves. Every edge in the figure connects a taxon to one of two group nodes, "Organic" or "Conventional," indicating that the taxon meets the prevalence and abundance criteria within that group. A taxon connected to both group nodes meets the criteria in both groups; a taxon connected to only one is specific to that group.

Interpretation

This is a valid way to visualize shared and group-specific "core" taxa, but it is a bipartite membership graph, not a network of relationships between taxa. Framing it as a "network" borrows visual language, nodes, edges, connected clusters, that in microbiome papers usually implies inferred ecological or statistical relationships between organisms. Here, the only relationship shown is "present above threshold in group X," which is a considerably simpler claim.

The "Core Microbiota" Claim

Observation

Requiring both a prevalence threshold (75% of replicates) and an abundance threshold (0.01%) to define "core" taxa is, on its own, a more defensible approach than using an abundance cutoff alone. Elsewhere in the paper's analysis pipeline (see Redo in R), the "core" dataset used for several other claims is filtered by abundance only, with no prevalence requirement at all. Figure 4's stricter definition is a genuine methodological improvement over that.

Interpretation

That said, "core microbiota" is a term that usually implies more than "present in most replicates and above a low abundance floor." It says nothing about function, about whether the taxon is stable in relative abundance (as opposed to just present), or about whether its presence is biologically meaningful versus incidental. Calling this the "core microbiota" risks the term doing more persuasive work than the underlying criteria justify, even with the improved prevalence filter.

Unknown

In the underlying OTU table, Erwinia occurs in exactly 1 of 48 samples, a single conventional calyx-end replicate, well below any reasonable prevalence threshold in any single tissue-treatment group. Erwinia's node in Figure 4 sits in a densely connected region of the graph, alongside Pantoea and Tatumella, where many nodes receive edges from both hub circles. Whether Erwinia specifically has an edge to only the Conventional hub, or to both, is difficult to determine with confidence from the figure alone at this density, and would need to be checked against the actual node/edge data behind the figure, if available, rather than by visual inspection.

Observation

Not every comparison points to a discrepancy. Deinococcus appears in the figure only on the organic-specific branch, which is consistent with the finding on Alpha Diversity that Deinococcus-Thermus is restricted to organic apple samples.

Claims About Harmful and Beneficial Taxa

Figure 4 places two taxa already discussed elsewhere on this site squarely on opposite, treatment-specific branches: Escherichia-Shigella under Conventional, and Lactobacillus under Organic. The paper's Discussion section extends this framing explicitly to other genera shown in the same figure:

In contrast, Ralstonia and Erwinia, frequently described for adverse impact on plant health (Denny, 2007; Pirhonen et al., 2018), prevailed in conventional apples. source: Wassermann et al. 2019, Discussion

Noteworthy among them are Methylobacterium, Hymenobacter, Spirosoma, and Zymomonas which were high abundant in organically managed apples, and Burkholderia, Pantoea, Erwinia, and Acinetobacter, especially high abundant in conventional apples. source: Wassermann et al. 2019, Results

Observation

Supplementary Table 1 gives specific numbers for this claim: 0.47% (Organic) vs. 1.26% (Conventional), marked as statistically significant. In the underlying OTU table used throughout this site, Erwinia occurs in exactly 1 of 48 samples total (see above), a single conventional calyx-end replicate, at 45 raw reads out of that sample's 428,514 clean reads, roughly 0.01% within that one sample. Averaged across all 24 conventional replicates, 23 of which are zero, this comes to approximately 0.0004% dataset-wide, over 1,000 times lower than the paper's reported 1.26%.

Interpretation

This is a large enough gap that it is difficult to attribute to normal pipeline variation alone. Even allowing for the roughly 30-fold difference in total OTU counts between this independent re-analysis and the original pipeline (see Data Preparation), no reasonable way of computing "abundance in the whole dataset" from a single 45-read observation in one sample reaches 1.26%. Either the two analyses are identifying substantially different sets of sequences as Erwinia, or something else in how the reported figure was calculated is not explained in the paper. "Prevailed" does not describe a taxon present in one sample out of forty-eight, whatever the correct dataset-wide percentage turns out to be.

Interpretation

A similar pattern holds for Ralstonia (see Microbiota Structure): average abundance is higher in conventional apples, but the single highest abundance found anywhere in the dataset, roughly 80%, occurs in one organic seed sample. Framing both genera as generally "prevailing" in conventional apples, when their apparent dominance is driven by a small number of individual samples rather than a consistent treatment-wide pattern, overstates what the replicate-level data show.

The Discussion frames organic-associated taxa in similarly strong terms:

Methylobacterium, identified to enhance the biosynthesis of strawberry flavor compounds (Verginer et al., 2010), was significantly higher abundant in organic apples [...] Controversially, Lactobacillus, which is frequently used within probiotics (Derrien and van Hylckama Vlieg, 2015), was one of the core taxa of organic apples. source: Wassermann et al. 2019, Discussion

Interpretation

As discussed on Microbiota Structure, Lactobacillus is very rare overall (0.2% total abundance), and its single highest reading in the entire dataset occurs in a conventional peel sample, not an organic one. Describing it as a "core taxon of organic apples" does not match the replicate-level data.