| CPC G16B 30/00 (2019.02) [C12Q 1/6881 (2013.01); G16B 20/00 (2019.02); G16B 40/20 (2019.02); G16B 50/10 (2019.02)] | 19 Claims |
|
1. A method comprising:
(a) obtaining a biological sample comprising a plurality of biological macromolecules from a plurality of distinct cell types from a subject having cancer;
(b) processing said biological sample to generate a feature profile of said plurality of biological macromolecules, wherein said feature profile comprises a plurality of features associated with said plurality of distinct cell types;
(c) computer processing said feature profile, using a deconvolution module, to (1) quantify an abundance of at least one of said plurality of distinct cell types in said biological sample, wherein quantifying said abundance comprises applying a batch correction procedure to remove technical variation in said abundance, and (2) generate one or more differential gene expression profiles that are differential across subtypes of said at least one of said plurality of distinct cell types,
wherein said batch correction procedure removes said technical variation between a reference matrix B of a plurality of feature signatures and said feature profile,
wherein said batch correction procedure is applied in a single cell reference mode (S-mode) or a bulk reference mode (B-mode),
(i) wherein said applying said batch correction procedure in said S-mode comprises removing technical differences between said reference matrix B derived from a set of single cell reference profiles and an input set of mixture samples M by:
(1) obtaining a plurality of estimates of a plurality of cell frequencies F* within said input set of mixture samples M, given said reference matrix B and said set of single cell reference profiles R, and
(2) refining said plurality of estimates of said plurality of cell frequencies F* by performing said batch correction procedure on said reference matrix B to obtain an adjusted reference matrix, and applying said adjusted reference matrix to said input set of mixture samples M, and
(ii) wherein said applying said batch correction procedure in said B-mode comprises removing said technical differences between said reference matrix B derived from bulk reference profiles and an input set of mixture samples M by:
(1) generating a plurality of mixture samples M* comprising a linear combination of a plurality of imputed cell type proportions in said input set of mixture samples M and corresponding profiles in said reference matrix B, and
(2) performing said batch correction on said input set of mixture samples M to eliminate said batch effects between said input set of mixture samples M and said plurality of mixture samples M*;
(d) predicting a clinical outcome of a cancer therapy on said subject for said cancer, based at least in part on said abundance and said one or more differential gene expression profiles of said at least one of said plurality of distinct cell types in said biological sample, wherein said at least one of said plurality of distinct cell types in said biological sample comprises a type of cancer cell; and
(e) administering said cancer therapy to said subject based on said predicted clinical outcome of said cancer therapy, wherein said cancer therapy is selected from the group consisting of a chemotherapy, an immunotherapy, and an immunochemotherapy.
|