bioRxiv2023
Construction of organ-specific gene expression datasets that include hundreds to thousands of experiments would greatly aid reconstruction of gene regulatory networks with organ-level spatial resolution. However, creating such datasets is greatly hampered by the requirements of extensive and tedious manual curation. Here we trained a supervised classification model that can accurately classify the organ-of-origin for a plant transcriptome. This K-Nearest Neighbor-based multiclass classifier was used to create organ-specific gene expression datasets for the leaf, root, shoot, flower, seed, seedling, silique, and stem in the model plant Arabidopsis thaliana. In the leaf, root, flower, seed and, a gene regulatory network (GRN) inference approach was used to determine: i. influential transcription factors (TFs) in that organ and, ii. the most influential TFs for specific biological processes in the organ. These genome-wide, organ-delimited GRNs (OD-GRNs), identified de novo many known regulators of organ development and processes operating in those organs. Moreover, many previously unknown TF regulators were highly ranked as potential master regulators of organ development or organ-specific processes. As a proof-of-concept, we focused on experimentally validating the predicted TF regulators of lipid biosynthesis in seeds, with relevance to food and biofuel production. Of the top twenty candidate TFs, eight (e.g., WRI1, LEC1, and FUS3) are known regulators of seed oil content. Importantly, we validated that seven more candidate TFs, whose role was previously unknown in seed lipid biosynthesis, indeed affect this process by genetics and physiological approaches, thus yielding a net accuracy rate of >75% for the de novo TF predictions. The general approach developed here could be extended to any species with sufficiently large gene expression datasets to speed up hypothesis generation and testing for constructing gene regulatory networks at a high spatial resolution. Significance StatementOur study develops a machine-learning framework for building extremely large gene expression datasets for each organ, and to infer organ-delimited gene regulatory networks. We show that this approach is very successful at predicting which transcription factors are going to regulate processes at an organ level. We validated the accuracy of the predictions for transcription factor regulators using the seed lipid synthesis pathway as a case study. We demonstrated a very high success rate for uncovering both known and novel transcription factor regulators for the seed lipid biosynthesis pathway. The approach described in this study is broadly applicable across any organism (plant or animal) that has a large body of public gene expression data.