S13 - Session P3 - Field multi-omics analysis using causal discovery method reveals agroecological interactions that improve vegetable yield and quality

Authors: Fuki Fujiwara *, Naoto Nihei, Atsushi Fukushima, Kenta Suzuki, Shohei Shimizu, Jun Kikuchi, Tomoko Matsumoto, Kae Miyazawa, Yasunori Ichihashi

Agroecosystem consists of a complex network of interactions between plants, microbes, and soils. Since this network includes many ecological functions to support plant growth, untangling this network will　contributes to the improvement of crop yield and quality without depending on chemical fertilizers. One promising approach for investigating this agroecosystem is network analysis using field multi-omics data consisting of omics datasets of crops, rhizosphere microbes, and soil components. However, the methods currently used for the network analysis of non-time series data often face the limitation of not being are able to distinguish between causal relationship and correlation. Here, we show that causal discovery method can be applied to field multi-omics data to reveal the interactions in the agroecosystem. Our statistical analysis combining a causal discovery model called Linear non-Gaussian Acyclic Model with other multi-omics data integration methods visualized a causal network associated with yield and quality of Brassica rapa planted in a field. The causal network showed that a trade-off between fresh weight and Brix value was controlled by the balance of carbon and nitrogen (C/N ratio) in the crop body. This is consistent with the fact that the increased nitrogen input to enhance yield can degrade the nutritional values of leafy vegetables. The causal network also showed that certain factors, such as rhizosphere methylotrophic bacteria and soil organic acids, positively affected fresh weight or Brix value without changing the C/N ratio. These findings suggest that manipulations of these factors possibly improve crop yield and nutritional quality beyond the trade-off caused by the nitrogen uptake. By focusing on causal discovery method, our study demonstrated the potential of field multi-omics data for engineering agroecosystem to improve crop performance.