C.Origami Deep Learning for Cell Type-Specific Chromatin Architecture Prediction

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Current methods for predicting cell type-specific chromatin folding are limited by their reliance on either DNA sequence features or chromatin data, failing to integrate both effectively and requiring large cell numbers, which is prohibitive for rare cell types.

Innovation Solution

C.Origami, a deep neural network, synergistically integrates DNA sequence features and cell type-specific genomic features such as DNA-binding protein profiles and chromatin accessibility information to accurately predict cell type-specific chromatin architecture.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If chromatin architecture capture technologies such as Hi-C are used to examine chromatin folding, then measurement precision of chromatin architecture is improved, but device complexity and cost increase, and large cell numbers are required

Engineering Contradiction:

Improvechromatin architecture measurement precisionVSAvoidexperimental complexity

Core Design Contradiction:

Measurement precisionVSDevice complexity

Solution Approach 1:

The patent replaces complex wet-lab chromatin architecture capture technologies (Hi-C, ChIP-seq, ATAC-seq) with a computational deep learning model (C.Origami) that processes DNA sequence and genomic feature data to predict chromatin architecture. This substitution eliminates the need for complex experimental procedures while maintaining prediction accuracy, directly resolving the contradiction between measurement precision and experimental complexity.

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

2Quantity of substance

If chromatin architecture capture technologies are used, then chromatin folding information is obtained, but large cell numbers are required which prohibits applications in rare cell types

Engineering Contradiction:

Improvecell number requirementVSAvoidchromatin architecture detection accuracy

Core Design Contradiction:

Quantity of substanceVSMeasurement precision

Solution Approach 1:

The patent creates a computational copy of chromatin architecture prediction through the C.Origami deep learning model, which simulates chromatin folding patterns based on DNA sequence and genomic features without requiring physical chromatin samples. This copying approach eliminates the need for large cell numbers while preserving the ability to accurately predict chromatin architecture in rare cell types where experimental methods fail.

Inventive Principle:

Principle #26Copying

3Adaptability or versatility

If approaches rely solely on DNA sequence for predictions, then simplicity is maintained, but cell type-specific chromatin interactions cannot be captured

Engineering Contradiction:

Improvecell type-specific prediction capabilityVSAvoidmodel complexity

Core Design Contradiction:

Adaptability or versatilityVSDevice complexity

Solution Approach 1:

The patent merges multiple data types (DNA sequence, DNA-binding protein profiles, chromatin accessibility information) into a unified deep learning model framework. This integration allows the model to capture cell type-specific chromatin interactions by combining intrinsic DNA sequence features with cell type-specific genomic information, resolving the contradiction between model simplicity and prediction adaptability across different cell types.

Inventive Principle:

Principle #5Merging (Combining)

Data Source

PatentUS20250182856A1Cell type-specific prediction of 3D chromatin architecture

Publication Date: 2025.06.05 NEW YORK UNIV

AI summary

The present disclosure relates to technologies for predicting genomic features, such as 3D genomic folding, in a target cell. Provided are methods for predicting genome structure and computer-implemented machines configured to predict genomic features. Wherein predicting 3D genomic features includes a neural network model architecture integrating (1) nucleotide-level DNA sequences, and (2) cell type-specific genomic features, wherein the cell type-specific genomic features comprise (i) genomic DNA-binding protein binding profile information, and (ii) chromatin accessibility information.