Inventors at Georgia Tech have developed a portable, robust and universal pattern classification technique capable of handling data acquired from multiple modalities that enables comparisons between different multi-cellular systems. Deriving physiologically meaningful metrics that can detect subtle differences between spatial phenotypes is a challenge; this innovation has developed an algorithm that is computationally tractable and uses spatially-derived networks, for data analysis via network theory. This method allows the extraction of global metrics such as path lengths and connectivity information as well as local metrics based on specific clusters of cell phenotypes. Sub-network identification and characterization allows for greater network quantification and enriches the overall metric space. This technology has been used to analyze several independent multicellular systems including: computational models of mouse embryonic stem cells (ESC) differentiation, spontaneous differentiation of ESC aggregates at both early and late stages, and gastrulation within cichlid fish. Furthermore, histological staining and confocal microscopy data sets have all been analyzed using the same algorithm demonstrating the broad applicability of this technology.
- Enables cross-system comparisons through quantitative analysis of multi-cellular patterns
- Flexibly handles multiple data input types such as histological staining, confocal imaging data, or simulated modeling outputs
- Applicable to many biological systems including stem cell differentiation and developmental biology investigations
- Automation of pathology analysis
- Investigating cell differentiation and “on-chip” organoid screening
- Confocal microscopy software suites for advanced in image analyses
The quantitative description of emergent spatial patterns in computational and morphogenic multicellular systems presents a significant challenge. In experimental systems, pattern identification of specific phenotypes is often done simply by visual inspection, which makes comparison with computational models difficult.