Systems and methods for deconvolution of expression data
Inventors
Zaitsev, Aleksandr • Chelushkin, Maksim • Cheremushkin, Ilya • Nuzhdina, Ekaterina • Zyrin, Vladimir • Dyikanov, Daniiar • Bagaev, Alexander • Ataullakhanov, Ravshan • Shpak, Boris
Assignees
Interested in licensing this patent?
MTEC can help explore whether this patent might be available for licensing for your application.
Publication Number
US-11587642-B2
Publication Date
2023-02-21
Expiration Date
Abstract
Techniques for determining one or more cell composition percentages from expression data. The techniques include obtaining expression data for a biological sample, the biological sample previously obtained from a subject, the expression data including first expression data associated with a first set of genes associated with a first cell type; determining a first cell composition percentage for the first cell type using the expression data and one or more non-linear regression models including a first non-linear regression model, wherein the first cell composition percentage indicates an estimated percentage of cells of the first cell type in the biological sample, wherein determining the first cell composition percentage for the first cell type comprises: processing the first expression data with the first non-linear regression model to determine the first cell composition percentage for the first cell type; and outputting the first cell composition percentage.
Core Innovation
The invention relates to a method and corresponding system for analyzing cancer biological samples by estimating a cell composition percentage for a selected cell type. RNA expression data for a biological sample from a subject having, suspected of having, or at risk of having cancer is obtained, where the RNA expression data includes first RNA expression data associated with a first set of genes associated with a first cell type.
For determining the first cell composition percentage, only the first RNA expression data is provided as input to a first non-linear regression model. The first non-linear regression model comprises a first ensemble of prediction models trained using gradient boosting, and the output represents an estimated percentage of RNA from the first cell type. The first cell composition percentage is then determined based on the estimated percentage of RNA from the first cell type.
The described framework further supports multiple cell types and cell-type refinement, including determining a second cell composition percentage from second RNA expression data using a second non-linear regression model, and simultaneously determining a plurality of cell composition percentages for a plurality of cell types using respective non-linear regression models. The document also describes determining a malignancy expression profile using an RNA expression profile and the first cell type’s cell composition percentage.
Claims Coverage
Independent claims are directed to a method, a system, and a non-transitory computer-readable storage medium. The core inventive coverage centers on estimating cell composition percentages from RNA expression data using a non-linear regression ensemble trained with gradient boosting and gene sets defined in Table 2, with dependent claims extending to staged sub-model estimation, multi-cell-type estimation, and malignancy expression profile determination.
Estimating cell composition percentage from cell-type gene set RNA expression
Obtaining RNA expression data for a biological sample from a subject having, suspected of having, or at risk of having cancer, where the RNA expression data includes first RNA expression data associated with a first set of genes associated with a first cell type, with the first RNA expression data consisting of expression data for at least 10 genes selected from the group of genes for the first cell type listed in Table 2, and determining a first cell composition percentage indicating an estimated percentage of cells of the first cell type in the biological sample.
Non-linear regression ensemble with gradient boosting using only cell-type RNA input
Determining the first cell composition percentage comprises providing only the first RNA expression data as input to a first non-linear regression model to obtain an output representing an estimated percentage of RNA from the first cell type, where the first non-linear regression model comprises a first ensemble of prediction models trained using gradient boosting.
Gene set defined by Table 2 for selected cell types
The first cell type is selected from the group consisting of B cells, CD4+ T cells, CD8+ T cells, endothelial cells, fibroblasts, lymphocytes, macrophages, monocytes, NK cells, neutrophils, and T cells, where Table 2 provides the group of genes for each cell type.
Malignancy expression profile based on estimated cell composition
The method further determines a malignancy expression profile using an RNA expression profile and the first cell type’s cell composition percentage.
The independent claims cover estimating cell composition percentages for a selected cell type in a cancer biological sample by using only the cell-type-associated RNA expression data as input to a non-linear regression ensemble trained with gradient boosting, with gene sets specified in Table 2 and with optional downstream determination of a malignancy expression profile using the estimated cell composition.
Stated Advantages
Not explicitly described in patent.
Documented Applications
Not explicitly described in patent.
Interested in licensing this patent?