Multimodal GiaC-phenomic approach for microbiome-tree system profiling

We constructed a computational methodology to assess health of plant-microbiome system through microbiome structure modelling combined with plant remote sensing. As a test dataset, we selected soil mycobiome and morphometry of Tilia cordata in nursery and forest sites. Our method is also applicable on forest or regional scale.

Microbiome part called GiaC ( G u i lds a nd o C currences) combines taxonomic and trophic composition as well as species co-occurrence modelled with advanced graph methods. We complemented state-of-the-art approaches with novel ones for visualisations, species filtering (Flexible99) and graph transformation modelling species clusters (ClusterCollapse). Flexible99 is a method that adjusts the species abundance cut-off to each sample set and removes rare species. ClusterCollapse generalises co-occurrence networks to species clusters by edge contraction and serves as an implicit homogeneity test.

To assess biomass of the seedlings we used low-cost and field-adopted morphometric and manual measurements. Top and side tree images, acquired with handheld RGB camera, were analysed using colour segmentation and pixel count based methods. Parameters, such as crown size, shape, area and pigment content, number of leaves, branch length and foliage density, allowed the seedlings to be classified into three different vitality groups.

Presented multimodal approach was capable to differentiate and characterize distinct best, suboptimal or critical states of microbiome-host system, both on microbial and plant side. Our results show that more stable fungal co-occurrence patterns should be attributed to the plant set of the best growth. In contrast, more chaotic patterns can be considered non-optimal for plant-mycobiome cooperation.