Fig 1.
Experimental design for maize.
Two generations (G0, G1) of maize experiment were grown in the greenhouse with microbiomes extracted from agricultural (MAg) and forest (MForest) soils. Generation 1 plants were grown with microbiomes that were direct descendants of those from generation 0. full-water (Ww) and half-water (Lw) treatments were imposed to plants during generation 0. The same water treatments were imposed in generation 1 after seed germination, but now so that for half of the plants in each treatment the water treatment was switched.
Fig 2.
Flowchart of the study design.
Schematic LDA topic modeling for K = 3 topics.
Fig 3.
Topic abundance weighting for the treatment and species types.
All topics to the left of topic 9 are associated with Acropora cervicornis species. All topics to the right of topic 8 are associated with Acropora palmata species. All topics below topic 11 are associated with Control treatment whereas all topics above topic 15 are associated with Inoculated (diseased or visually unaffected) treatment. Topics in color are topics associated with an experimental outcome. Topic 20 is associated with Low disease susceptibility; topics 5 and 14 are associated with Medium disease susceptibility; topics marked with triangle up symbols are associated with High disease susceptibility.
Fig 4.
Ternary plot of topic abundances for the outcome type (control, diseased, visually unaffected).
Fig 5.
Distribution of ASV sequences in each learned LDA topic.
Topics to the left of topic 9 are strongly associated with Acropora cervicornis species, and topics to the right of topic 8 are strongly associated with Acropora palmata species. The labels are written in the phylum_class_order_family format, the names of genera, species, and ASVs are not shown. Only probabilities (color circles) greater than 0.1 are shown. Smaller circles are displayed on the top of larger circles.
Table 1.
Comparison of LDA results and results obtained from the relative abundance analysis (Tables 2 and 3 in [19]).
Table shows mean and standard deviation (SD) relative abundances (RA) in % by taxon of core microbiomes per experimental outcome and coral-host [19] as wells as probability % of these taxa in LDA topics associated with corresponding outcome and coral species. Multiple LDA topics were associated with Diseased outcome in both coral species. Topic 20 and topic 5 were associated with low and medium disease susceptibility, respectively, whereas topics 19, 1, 6, and 18 were associated with high disease susceptibility.
Table 2.
Number of topics associated with experimental conditions at each taxonomic level.
Strong (topic abundance is four times (×4) larger than for other treatments) and moderate (topic abundance is twice (×2) as large but less than four times larger than that for other treatments) topic associations are shown for each experimental condition. The numbers in parentheses represent number of topics associated with at least one experimental condition and overall number of topics used in LDA at taxonomic level.
Table 3.
Number of topics at each taxonomic level that showed statistically significant relationships between topics and plant traits based on a Spearman’s rank correlation coefficient with Holm–Bonferroni correction.
The numbers in parentheses represent number of different topics associated with significant relationships and overall number of topics used in LDA at taxonomic level. The last four columns are Leaf mass per area, % Leaf water content, Stomatal conductance, and Water use efficiency.
Table 4.
Most probable and amplified taxa in some topics associated with experimental conditions at phylum, class, and order taxonomic levels.
Left: average percentages of most abundant taxa in pots for each taxonomic level. Middle: topic association with experimental conditions (HW: half-water, FW: full-water, G0: generation 0, G1: generation 1, AG: agricultural soil source). Right: most probable taxa in topic with probabilities are displayed in each upper row; most amplified taxa with its value (see Methods) are displayed in each lower row, some taxa were unidentified at lower levels. Note that unidentified taxa were aggregated to the highest known taxonomic level, meaning that they may represent one or several classes, orders, etc. Not all topics are listed.
Table 5.
Most probable and amplified taxa in topics strongly associated with experimental conditions at family and ASV taxonomic levels.
Left: average percentages of most abundant taxa in pots for each taxonomic level. Middle: topic association with experimental conditions (HW: half-water, FW: full-water, G0: generation 0, G1: generation 1, AG: agricultural soil source, FR: forest soil source, ST: stable watering, SW: switched watering). Right: most probable taxa in topic with probabilities are displayed in each upper row; most amplified taxa with its value (see Methods) are displayed in each lower row, some taxa were unidentified at lower levels. Note that unidentified taxa were aggregated to the highest known taxonomic level, meaning that they may represent one or several classes, orders, etc. Not all topics are listed.
Fig 6.
Topic abundance weighting for the water treatment and generation on a 0–1 scale.
All topics to the left of topic 5 are associated with generation 0. All topics to the right of topic 25 are associated with generation 1. All topics below topic 25 are associated with half-water treatment whereas all topics above topic 15 are associated with full-water treatment. Topics are colored according to the soil microbiome inoculation source. The shapes of the markers corresponding to the topic association with stability of watering treatment. Note: the weighting was calculated by how much average topic abundance under one treatment was different from that under alternative one (see Methods).
Fig 7.
Distribution of ASVs in each learned LDA topic.
Topics to the left of topic 10 are associated with half-water treatment, and topics to the right of topic 15 are associated with full-water treatment. The labels are written in the phylum_class_order_family_genus_species format, the names of genera, species, and ASV are not shown. Only probabilities (color circles) greater than 0.025 are shown. The sizes of the circles representing probabilities are multiplied by 4 for visualization purposes. Smaller circles are displayed on the top of larger circles.
Fig 8.
Difference in abundances of ASVs in the half-water treatment relative to the full-water (dashed middle line).
Dots represent the differential abundance coefficient and the error bars are standard errors. The taxa shown are only those that are significant after a p-value correction with the FDR set to 0.05. Plots were produced using corncob R package [50]. The dots colored in red show taxa that were abundant in half-water related LDA topics. The dots colored in blue show taxa that were abundant in full water related LDA topics. Note that the comparison was made with the Fig 7 where only ASVs with p > 0.025 are shown. The second half of the plot with other significant taxa is shown in S12 Fig.