UniCell Deconvolve: Cell Type Deconvolution For Transcriptomic Data

UniCell Deconvolve (UCD) is a pre-trained deep learning model that provides context-free estimations of cell type fractions from whole transcriptome expression data for bulk, single-cell and spatial transcriptomics data. The model is trained on the world’s largest fully-integrated scRNA-Seq training database, comprising 28M+ single cells spanning 840+ cell types from 899 studies to date. Extensive benchmarking shows UCD favors comperably when compared with reference-based deconvolution tools, without the need for pretraining. UCD demonstrates strong multi-task performance across a range of deconvolution challenges spanning several transcriptomic data modalities, disease types, and tissues.

• Read our paper at Nature Communications.

• Install via pip install ucdeconvolve.

• Discuss development on GitHub.