Machine Learning Model for Transcriptomic Profiles from Omics Imaging
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current spatial transcriptomics technologies have limitations in providing a detailed understanding of complex tissues, such as the central nervous system, as they often rely on generalized large morphological features, failing to capture the underlying molecular consequences of patterns over large spatial areas effectively.
Innovation Solution
A computer-implemented method and system that utilizes machine learning to generate gene expression profiles from omics imaging data, including histology and spatial omics data, by processing images with a deployed machine learning model to produce transcriptomic profiles, enabling high-resolution spatial analysis and cell typing.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If spatial transcriptomics uses large morphological features (100 μm features), then the device complexity is reduced and ease of operation is improved, but the measurement precision and ability to capture molecular consequences at cellular resolution deteriorates
Solution Approach 1:
The patent segments the tissue analysis into multiple hierarchical levels: first analyzing large morphological features at 100 μm scale, then progressively subdividing into smaller regions to achieve cellular resolution (2 μm scale). This segmentation allows the system to manage complexity by breaking down the large-scale analysis into manageable smaller units while maintaining high measurement precision at the finest scale.
Solution Approach 2:
The patent introduces a spatial dimension hierarchy by operating at multiple scales (100 μm to 2 μm). This dimensional approach allows the system to simultaneously capture both large-scale tissue architecture and fine-scale cellular molecular profiles, resolving the contradiction between ease of operation at large scales and measurement precision at small scales.
2Area of stationary object
If spatial transcriptomics focuses on large spatial areas, then the area of analysis is increased and productivity is improved, but the measurement precision of molecular patterns at cellular level deteriorates
Solution Approach 1:
The patent divides the large spatial area into hierarchical segments: first processing broad tissue regions, then subdividing into smaller sub-regions for detailed molecular analysis. This segmentation enables the system to maintain high spatial coverage while achieving cellular-level molecular pattern resolution in each segment through progressive refinement.
Solution Approach 2:
The patent applies partial action by focusing high-resolution molecular analysis only on specific sub-regions of interest within the larger tissue area. Rather than applying maximum resolution uniformly across the entire tissue section, the system selectively intensifies analysis in key areas, maintaining overall productivity while achieving high measurement precision where needed.
Data Source
AI summary
The present disclosure relates to systems and method of determining transcriptomic profile from omics imaging data. The systems and methods train machine learning methods with intrinsic and extrinsic features of a cell and/or tissue to define transcriptomic profiles of the cell and/or tissue. Applicants utilize a convolutional autoencoder to define cell subtypes from images of the cells.