Systems and methods for dissecting heterogeneous cell populations

A method for single cell analysis comprising:

• under control of a hardware processor:
• receiving single cell gene expression data comprising a plurality of normalized reference gene expression vectors and a plurality of normalized test gene expression vectors representing single cell gene expression profiles of a plurality of reference cells and a plurality of test cells respectively, in a gene expression space;
• determining (i) a gene feature matrix, comprising a plurality of gene features, and, for each of the plurality of normalized reference gene expression vectors and the plurality of normalized test gene expression vectors, (ii) a reference gene feature weighting vector or a test gene feature weighting vector using a training matrix comprising a plurality normalized training gene expression vectors, wherein the plurality of reference gene feature weighting vectors and the plurality of test gene feature weighting vectors represent gene feature profiles of the plurality of reference cells and the plurality of test cells in a gene feature space, respectively;
• determining (i) a reference Gaussian mixture model comprising a plurality of reference Gaussian densities each associated with a reference weight and (ii) a test Gaussian mixture model comprising a plurality of test Gaussian densities each associated with a test weight from the plurality of reference gene feature weighting vectors and the plurality of test gene expression vectors, respectively, wherein each of (i) the plurality of reference Gaussian densities and associated reference weights and (ii) the plurality of test Gaussian densities and the associated test weights represents a subpopulation of the plurality of reference cells or the plurality of test cells, respectively;
• and
• determining a difference between the plurality of test cells, or a subpopulation thereof, and the plurality of reference cells, or a subpopulation thereof, using the plurality of reference Gaussian densities and the plurality of test Gaussian densities.

A system for single cell analysis comprising:

• non-transitory memory configured to store executable instructions and single cell gene expression data comprising a plurality of reference gene expression vectors and a plurality of test gene expression vectors representing single cell gene expression profiles of a plurality of reference cells and a plurality of test cells respectively, in a gene expression space;
• and
• a hardware processor in communication with the non-transitory memory, the hardware processor is programmed by the executable instructions to:
• for each of the plurality of reference gene expression vectors and the plurality of test gene expression vectors, normalize gene expression values of the reference gene expression vector or the test gene expression vector to generate a normalized reference gene expression vector or a normalized test gene expression vector comprising normalized gene expression values;
• determine (i) a gene feature matrix, comprising a plurality of gene features, and, for each of the plurality of normalized reference gene expression vectors and the plurality of normalized test gene expression vectors, (ii) reference gene feature weighting vector or a test gene feature weighting vector using a training matrix comprising a plurality normalized training gene expression vectors, wherein the plurality of reference gene feature weighting vectors and the plurality of test gene feature weighting vectors represent gene feature profiles of the plurality of reference cells and the plurality of test cells in a gene feature space, respectively;
• determine (i) a reference Gaussian mixture model comprising a plurality of reference Gaussian densities each associated with a reference weight and (ii) a test Gaussian mixture model comprising a plurality of test Gaussian densities each associated with a test weight from the plurality of reference gene feature weighting vectors and the plurality of test gene expression vectors, respectively, wherein each of (i) the plurality of reference Gaussian densities and associated reference weights and (ii) the plurality of test Gaussian densities and the associated test weights represents a subpopulation of the plurality of reference cells or the plurality of test cells, respectively;
• aligning each of the plurality of test Gaussian densities to an aligned reference Gaussian density of the plurality of reference Gaussian densities;
• and
• performing an analysis of the plurality of test cells relative to the plurality of reference cells using one, at least one, or each, of the plurality of test Gaussian densities and the aligned reference Gaussian density the test Gaussian density is aligned to.

A system for single cell analysis comprising:

• non-transitory memory configured to store executable instructions and single cell gene expression data comprising a plurality of normalized reference gene expression vectors and a plurality of normalized test gene expression vectors representing single cell gene expression profiles of a plurality of reference cells and a plurality of test cells respectively, in a gene expression space;
• and
• a hardware processor in communication with the non-transitory memory, the hardware processor is programmed by the executable instructions to:
• determine (i) a gene feature matrix, comprising a plurality of gene features, and, for each of the plurality of normalized reference gene expression vectors and the plurality of normalized test gene expression vectors, (ii) reference gene feature weighting vector or a test gene feature weighting vector using a training matrix comprising a plurality normalized training gene expression vectors, wherein the plurality of reference gene feature weighting vectors and the plurality of test gene feature weighting vectors represent gene feature profiles of the plurality of reference cells and the plurality of test cells in a gene feature space, respectively;
• determine (i) a reference Gaussian mixture model comprising a plurality of reference Gaussian densities each associated with a reference weight and (ii) a test Gaussian mixture model comprising a plurality of test Gaussian densities each associated with a test weight from the plurality of reference gene feature weighting vectors and the plurality of test gene expression vectors, respectively, wherein each of (i) the plurality of reference Gaussian densities and associated reference weights and (ii) the plurality of test Gaussian densities and the associated test weights represents a subpopulation of the plurality of reference cells or the plurality of test cells, respectively;
• align each of the plurality of test Gaussian densities to an aligned reference Gaussian density of the plurality of reference Gaussian densities;
• determine a Gaussian density difference between one of the plurality of test Gaussian densities and the aligned reference Gaussian density the test Gaussian density is aligned to;
• and
• determine a difference between the plurality of test cells, or a subpopulation thereof, and the plurality of reference cells, or a subpopulation thereof, using the Gaussian density difference.