Biogeochemical typing of paddy field by a data-driven approach revealing sub-systems within a complex environment--a pipeline to filtrate, organize and frame massive dataset from multi-omics analyses

Abstract

We propose the technique of biogeochemical typing (BGC typing) as a novel methodology to set forth the sub-systems of organismal communities associated to the correlated chemical profiles working within a larger complex environment. Given the intricate characteristic of both organismal and chemical consortia inherent to the nature, many environmental studies employ the holistic approach of multi-omics analyses undermining as much information as possible. Due to the massive amount of data produced applying multi-omics analyses, the results are hard to visualize and to process. The BGC typing analysis is a pipeline built using integrative statistical analysis that can treat such huge datasets filtering, organizing and framing the information based on the strength of the various mutual trends of the organismal and chemical fluctuations occurring simultaneously in the environment. To test our technique of BGC typing, we choose a rich environment abounding in chemical nutrients and organismal diversity: the surficial freshwater from Japanese paddy fields and surrounding waters. To identify the community consortia profile we employed metagenomics as high throughput sequencing (HTS) for the fragments amplified from Archaea rRNA, universal 16S rRNA and 18S rRNA; to assess the elemental content we employed ionomics by inductively coupled plasma optical emission spectroscopy (ICP-OES); and for the organic chemical profile, metabolomics employing both Fourier transformed infrared (FT-IR) spectroscopy and proton nuclear magnetic resonance (1H-NMR) all these analyses comprised our multi-omics dataset. The similar trends between the community consortia against the chemical profiles were connected through correlation. The result was then filtered, organized and framed according to correlation strengths and peculiarities. The output gave us four BGC types displaying uniqueness in community and chemical distribution, diversity and richness. We conclude therefore that the BGC typing is a successful technique for elucidating the sub-systems of organismal communities with associated chemical profiles in complex ecosystems.

Figure 1. Schematic representation for the Biogeochemical Typing (BGC typing).

Yellow box: steps for collection and pre-processing the samples. Orange box: steps for data acquisition and formatting for BGC typing. Red box: steps for BGC typing as the integrated statistical analyses.

Figure 2. Schematic representation for the sampling location.

Shadow map showing Kanto region (Japan). Grey area is Saitama prefecture. The rectangle indicates where the samples were taken. Magnified area is schematic map for the sampling site. At right side, schematic figure: Ara River was sampled in two points (ara1 and ara2) as well as paddy fields located within the area between these two river sampling points. Three independent paddy fields (paddy field 1, 2, 3) were selected and three samples were taken from each paddy field (p1f1-3, p2f1-3 and p3f1-3). These paddy fields were connected with Ara River through independent collector streams which were also sampled (p1s, p2s, p3s), respectively for each paddy field. Blue arrows indicate flow direction of Ara river and light brown arrows indicate each sampling points. Gaps in Ara River indicate bridges.

Figure 3. Chemical profiles for the sampling points.

A) ICP-OES heatmap. X-axis: elements. Y-axis: sampling points. Red colour intensity corresponds to elemental concentration normalized by element. B) FT-IR spectra. X-axis: wavelength number. Y-axis: absorbance intensity. C) ¹H-NMR spectra. X-axis: chemical shift. Y-axis: intensity.

Figure 4. Finding the optimal number of BGC types.

X-axis: number of clusters. Y-axis: sum of squared distances from each variable to the centroid within the BGC type.

Figure 5. Delimiting BGC types.

Plot of PCA scores for the extracted OTU matrix correlated with chemical profile. Four BGC types were delimited by k-means clustering. X-axis: PC1. Y-axis: PC2. BGC I: area enclosed in red with cross symbols, BGC II: area enclosed in blue with x symbols, BGC III: area enclosed in green with circular signals, BGC IV: area enclosed in pink with triangular signals. Arrows indicate the axes separating the BGC types; the quasi-horizontal arrow separates BGC III from BGC IV along the PC1 axis; the quasi-vertical arrow separates BGC I from BGC II along the PC2.

Figure 6. Community distributions of the BGC types over the sampling points.

Distributions of relative abundances of communities identified as Archaea, 16S rRNA and 18S rRNA for the BGC types along the sampling points. X-axis: sampling points, Y-axis: relative abundance. Green shadows indicate sampling points on lentic waters and blue shadows indicate the sampling points over lotic waters. Below: a schematic drawing for the sampling points.

Figure 7. Heatmap for the Spearman correlations of extracted OTUs against chemical profiles.

X-axis, in order: chemical elements (ICP-OES), wavelength number (FT-IR), chemical shifts (H-NMR). Y-axis: OTUs arranged from BGC type I (upper) to IV (bottom). BGC types are indicated on the left. The meaning of the colours is indicated in the legend at the bottom.

Figure 8. Comparison of the average correlations for each BGC type against PCA loadings.

X-axis in order: chemical elements (ICP-OES), wavelength number (FT-IR), chemical shifts (H-NMR) (legend omitted). Y-axis: scaled scores. Average correlations from top to bottom are cluster 1 against PC2, cluster 2 against PC2, cluster 3 against PC1, and cluster 4 against PC1. Values are scaled by unity of variance.