DISSECT: deep semi-supervised consistency regularization for accurate cell type fraction and gene expression estimation

TY - JOUR

T1 - DISSECT: deep semi-supervised consistency regularization for accurate cell type fraction and gene expression estimation

AU - Khatri, Robin

AU - Machart, Pierre

AU - Bonn, Stefan

PY - 2024/4/30

Y1 - 2024/4/30

N2 - Cell deconvolution is the estimation of cell type fractions and cell type-specific gene expression from mixed data. An unmet challenge in cell deconvolution is the scarcity of realistic training data and the domain shift often observed in synthetic training data. Here, we show that two novel deep neural networks with simultaneous consistency regularization of the target and training domains significantly improve deconvolution performance. Our algorithm, DISSECT, outperforms competing algorithms in cell fraction and gene expression estimation by up to 14 percentage points. DISSECT can be easily adapted to other biomedical data types, as exemplified by our proteomic deconvolution experiments.

AB - Cell deconvolution is the estimation of cell type fractions and cell type-specific gene expression from mixed data. An unmet challenge in cell deconvolution is the scarcity of realistic training data and the domain shift often observed in synthetic training data. Here, we show that two novel deep neural networks with simultaneous consistency regularization of the target and training domains significantly improve deconvolution performance. Our algorithm, DISSECT, outperforms competing algorithms in cell fraction and gene expression estimation by up to 14 percentage points. DISSECT can be easily adapted to other biomedical data types, as exemplified by our proteomic deconvolution experiments.

U2 - 10.1186/s13059-024-03251-5

DO - 10.1186/s13059-024-03251-5

M3 - SCORING: Journal article

C2 - 38689377

VL - 25

SP - 112

JO - GENOME BIOL

JF - GENOME BIOL

SN - 1474-760X

IS - 1

M1 - 112

ER -