artiMATE - artificial neural networks of satellite imagery, fungal metagenomes and high-throughput interaction studies

artiMATE -  artificial neural networks of satellite imagery, fungal metagenomes and highthroughput interaction studies.
    
Machine learning (ML) has been extensively adopted to medical and environmental image analysis to classify image features related to disease and agriculture, respectively. However, in ML related to satellite image analysis for climate and agricultural studies there is often a scarcity of high-quality annotated data, which combined with limited understanding of the networks comprising soil microbiomes, limits the adoption of ML for crop protection. In this project we aim to combine a small number of ground truth labeled images from global DNA-annotated soil microbiome sampling sites for semi-supervised and active learning to build a model of two challenging learning tasks in agriculture, namely the prediction of i) spatio-temporal patterns in satellite imagery of soil microbiome sampling sites across planet Earth, and ii) the causal interactions between microbe sustaining soil microbiomes of various categorized environments inferred from DNA barcodes. Acknowledging the importance of soil microbiomes for crop yields, this project will furthermore experimentally test modeled signaling modalities (i.e. receptors, signal peptides) in selected microbes using a newly developed high-throughput (HTP) functional assay connecting receptors with signaling chemistry. The project thus aims to vastly extend our understanding of the causal factors underlying microbial networks in diverse soil environments with the ultimate goal to support sustainable smart farming.