Computational model for predicting gene expression profiles across human cell types

This technology is a machine learning model that predicts gene expression across human cell types by learning regulatory grammar from genomic sequence context and chromatin accessibility data.

Unmet Need: Predicting cell-type-specific gene expression from regulatory features

Transcriptional regulation governs the cell-type-specific gene expression programs that underlie human development and genetic disease. Predicting transcriptional changes has broad applications in experimental design, the interpretation of regulatory variation, and the understanding of disease mechanisms. However, existing computational approaches to predict gene expression are often trained on limited datasets or specific cell types and therefore struggle to generalize across diverse cellular contexts. In addition, although large-scale datasets capturing regulatory features such as chromatin accessibility and transcription factor binding motifs are increasingly available, current models do not fully leverage these data to infer transcriptional regulation across cell types.

The Technology: Computational pipeline to predict gene expression across diverse cell types

This technology describes a computational pipeline that predicts gene expression across diverse human cell types. The model is trained using publicly available regulatory (e.g., chromatin accessibility, transcription factor binding motifs, etc.) and gene expression data. It can accurately predict gene expression even for cell types not included in the training set. This technology can be used to identify regulatory modules controlling the expression of target genes and predict transcriptional changes resulting from regulatory perturbations such as substitutions, deletions, and insertions. As such, this technology has broad applications in experimental design, the interpretation of regulatory variation, and the understanding of disease mechanisms.

Applications:

• Gene expression prediction across diverse human cell types
• Identification of cis-regulatory elements and regulatory modules
• Prioritization of candidate regulatory elements for experimental validation
• CRISPR-based therapy response prediction
• In silico perturbation of regulatory elements
• Non-coding mutation effect evaluation
• Enhancer-promoter target prediction
• Transcription factor interaction analysis
• Generation of databases of gene editing sites

Advantages:

• Generalizes gene expression prediction to cell types not included in training
• Integrates multiple regulatory features
• Enables in silico analysis of regulatory region perturbations
• Identifies regulatory modules controlling target gene expression
• Trained with reliable, rich multi-omics data sets
• Enables in silico CRISPR
• Data-driven prediction model
• Cost-effective CRISPR target selection

Lead Inventor:

Raul Rabadan, Ph.D.

Patent Information:

Patent Pending (US20260051366)

Related Publications:

• Fu X, Mo S, Buendia A, et al. A foundation model of transcription across human cell types. Nature. 2025;637(8047):965-973.

Tech Ventures Reference:

• IR CU23168

• Licensing Contact: Joan Martinez