High-throughput in-silico promoter variant scoring for plant gene expression modulation

Overview

This solution presents an advanced in-silico pipeline to predict and rank single-nucleotide variants (SNVs) in plant promoters or regulatory elements, modulating gene expression. Using rice as a model, this species-agnostic approach integrates promoter and transcription start site (TSS) annotations, transcription factor (TF) binding models, and chromatin accessibility data to train deep learning models. The pipeline aims to provide gene-wise shortlists with predicted fold-changes in expression and potential mechanisms, facilitating cost-effective and efficient exploration of genetic variation before laboratory validation.

Technical specifications

Key features:

• Utilizes an in-silico saturation mutagenesis approach around TSS to predict the direction and magnitude of expression changes per SNV.
• Incorporates an EMS-feasibility score and checks against natural variation databases like 3K rice to prioritize realistic and lower-risk alleles.
• Outputs include a containerized pipeline (e.g., Docker) with comprehensive code, models, and documentation, ready for deployment and further development.
• Leverages multiple genomic resources, including plant MPRAs, ATAC-seq regulome maps, and TF-binding datasets, to enhance prediction accuracy.

Technology readiness level

Currently, the technology is at TRL 3, indicating that it has been proven in a controlled research environment. The project includes a structured plan to advance the technology to TRL 4 through specific milestones such as data curation, model training, variant scoring, and mutagenesis suitability assessments, with a projected timeline of 12 months for completion.