Cytometry Data Analysis Using Marker Ratios for Biological Interpretation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for analyzing cytometry data face challenges such as the 'curse of dimensionality' due to high-dimensional data, loss of single cell information in clustering, and difficulty in interpreting results in a biological context.
Innovation Solution
A method that involves binning detected intensities of markers from cytometric measurements, generating marker functions such as ratios of intensities, and creating feature sets to identify pairs that exhibit the largest variation between two samples, facilitating classification and interpretation using machine learning techniques.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If dimensionality reduction techniques such as tSNE are used to analyze cytometry data, then the analysis complexity is reduced, but the results become difficult to interpret in a biological sense due to artificial axes
Solution Approach 1:
The patent transforms the high-dimensional cytometry data into a two-dimensional representation that preserves biological interpretability. By mapping marker intensities to x and y axes while using color coding for a third marker, the method reduces dimensionality without creating artificial axes, allowing direct biological interpretation of the visualized data.
Solution Approach 2:
The patent creates a visual copy of the high-dimensional data space in a two-dimensional plot. Each cell is represented as a point in 2D space with its position determined by two markers and its color indicating the intensity of a third marker, thereby copying the essential information in an interpretable format.
2Device complexity
If clustering methods such as PhenoGraph or CITRUS are used to analyze cytometry data, then the dimensionality is reduced, but single cell information is lost and clustering resolution is limited
Solution Approach 1:
Instead of fully clustering the data which would group cells and lose individual information, the patent applies a partial approach by visualizing each cell individually in 2D space. This allows dimensionality reduction while preserving single-cell resolution and information, avoiding the excessive action of complete clustering.
Solution Approach 2:
The patent segments the high-dimensional data space into a two-dimensional visualization where each cell remains as an individual entity. By dividing the complexity of high-dimensional analysis into a manageable 2D representation, the method maintains single-cell information while reducing analytical complexity.
3Measurement precision
If the number of fluorescent, mass or oligo markers is increased to improve cell differentiation, then the measurement precision is improved, but the analysis difficulty increases dramatically due to the huge number of possible combinations
Solution Approach 1:
The patent extracts the essential information from high-dimensional marker data by selecting and visualizing only the most relevant combinations of markers in a 2D plot. This extraction approach allows using multiple markers for precise cell differentiation while avoiding the complexity of analyzing all possible marker combinations.
Solution Approach 2:
By transforming multiple marker dimensions into a two-dimensional visualization with color coding, the patent reduces the complexity of analyzing numerous marker combinations. The method projects high-dimensional marker space into 2D while preserving differentiation capability through strategic marker selection and color-coded representation.
Data Source
AI summary
The invention relates to a method for classifying selected marker signals from cytometric measurements comprising a first measurement and a second measurement, wherein the first measurement comprises a cytometric measurement acquired from a first sample of particles, and the second measurement comprises a cytometric measurement acquired from a second sample of N2 particles, being the number of particles in the first sample and N2 being the number of particles in the second sample. Each particle njN 1 of the N1 particles of the first sample is labelled with a number of L1 fluorescent, mass or oligo markers ljN 1. Each particle n1N 2 of the N2 particles of second sample is labelled with a number of L2 fluorescent, mass or oligo markers ljN2′ With the acquired data a machine learning method is trained such that the marker combinations showing the most significant differences for two distinct populations are selected and displayed to a user in a novel fashion.