Gene expression prediction using biophysical model-informed machine learning
A novel in silico framework combining convolutional neural networks and biophysical models to predict gene expression from promoter sequences, facilitating high-throughput promoter screening and de novo design.
Overview
The proposed solution is an innovative in silico framework designed to predict gene expression from promoter sequences. This framework integrates a convolutional neural network (CNN) with a biophysical model, enabling accurate prediction of gene expression at single-cell resolution. The approach leverages genome-wide correlations and biophysical principles to offer a robust predictive tool that supports high-throughput promoter screening and de novo promoter design using advanced Bayesian optimization techniques.
Technical specifications
Key features:
- Convolutional Neural Network (CNN): Encodes promoter sequences and extracts intrinsic regulatory features, translating them into kinetic parameters.
- Biophysical Model: Provides a principled framework for capturing transcriptional dynamics, yielding transparent and accurate predictions.
- High-Throughput Screening: Capable of supporting mutagenesis assays and promoter variant screening to link promoter sequences with biophysical parameters.
- Bayesian Optimization: Facilitates the de novo design of promoters, enhancing commercial applications.
Technology readiness level
The technology is at a readiness level of 4. This indicates that the components of the solution have been validated in a laboratory setting, and the framework is in the stage of further development and testing through mutagenesis assays and CNN training.
About Queen's University
Queen’s University is a comprehensive public research institution with a large student body and global alumni network based in Kingston, Ontario. For industry, it offers a research and technology park with wet/dry labs, shared core facilities, and collaboration models from sponsored research to contract testing. Clinical translation is enabled by integration with Kingston’s major hospital system, providing governed access to patients, clinicians, and health data. Talent pipelines are built through structured internship programs and project-based engagements. Research is backed by Canada’s Tri‑Council agencies and the Canada Foundation for Innovation, plus provincial and industry support, and a dedicated tech transfer office advances IP, licensing, and startups.
delivered to your inbox.