Clustered Gene Expression Analysis for Mixed-Subtype DEG Detection

The patent segments the cell population into multiple subpopulations based on their distinct gene expression patterns. By identifying and separating cells into different clusters (subtypes) before DEG analysis, the method can accurately detect DEGs specific to each subtype rather than averaging across heterogeneous populations. This segmentation resolves the contradiction by enabling reliable DEG detection in mixed populations while maintaining manageable complexity through systematic clustering approaches.

The patent changes the statistical modeling approach from assuming unimodal distribution to accommodating multimodal distributions. By introducing parameters that allow for multiple peaks in the gene expression distribution (such as mixture models with multiple components), the method can capture the heterogeneity of cell populations. This parameter change enables accurate DEG detection while the structured implementation keeps the complexity controlled.

The patent performs preliminary clustering of cells into subpopulations based on overall gene expression patterns before conducting the probability distribution search for DEG analysis. This preliminary action simplifies the subsequent statistical modeling by reducing the complexity of the multimodal distribution into several simpler unimodal distributions, each representing a cell subtype. This approach improves measurement precision while controlling computational complexity through the two-stage process.

The patent segments the complex multimodal distribution into multiple simpler unimodal distributions, each representing a distinct cell subtype. By fitting separate probability distributions to each segmented subgroup rather than attempting to fit a single complex multimodal distribution to the entire population, the method achieves accurate measurement while reducing computational complexity through divide-and-conquer strategy.