EXTENSION OF TRANSCRIPTIONAL REGULATOR RESOURCE IN CASSAVA BY MACHINE LEARNING BASED MODELING

Transcriptional regulatory system of plant species is believed to be rather complicated and contains specific elements required for sessile organisms to survive under wide range of environmental conditions. The current findings of transcriptional regulatory cascade are always underestimated its complexity due to a lack of knowledge on the interactions of transcription factor proteins (TF) and the regulatory motifs in DNA, called TF binding sites (TFBS). This circumstance restricts our understanding into the response of plants to a particular exposed perturbation, which is a key prerequisite in the study pathway toward climateadaptive plantations. To alleviate this limitation, Plant-DTI (Plant DBD-TFBS Interaction) is previously developed as a cutting-edge machine learning based tool for TF-TFBS prediction. The model was constructed from at least 1,241,314 interacting pairs of DNA binding domains (DBD) and TFBSs from experiments. The data coverage allows Plant-DTI to be able to predict gene targets of TFs up to half of the DBD types existing in plant species (30 TF families and 336 TFBSs). In this study, Plant-DTI model was exploited to study transcriptional regulation of sucrose synthase genes (SUS), a sucrolytic enzyme involving in carbon allocation and root biomass synthesis in cassava. There are 150 putative TFs were predicted for SUS1 gene, the majority of which are related to stress-response. The prediction was found consistent with the previous yeast one-hybrid (Y1H) study. These putative regulators enable us to connect the transcriptional response of SUS1 genes as well as their function to the exposed environments, and subsequently to better dealing with the surrounding influence. Taken together, this study demonstrates the advantage of big data and computational modeling to facilitate our moving towards the precise management in cassava farming.