Fig. 3: scPRINT GN inference performance on cell-type specific ground truths.
From: scPRINT: pre-training on 50 million cells allows robust gene network predictions
A The ground truths are generated via orthogonal sequencing assays on the same cell type. ChIP-seq and perturb-seq are intersected for the MCalla et al. dataset on human (hESCs) and mouse (mESCs) Embryonic Stem Cells, whereas perturb-seq on the K562 cell line is only used for the genome-wide perturb-seq ground truth. B Performance of scPRINT, scPRINT (omnipath’s heads): same scPRINT version but with attention heads selected using a subset of omnipath, scPRINT (Han et al.’s heads): same scPRINT version but with attention heads selected using a subset of the Han et al.’s ground truth dataset, compared to GENIE3, DeepSEM, Geneformer v2, and scGPT on the MCalla et al. ground truth using the AUPRC and EPR on two human and two mouse ESC datasets. C Same as (B) but on the genome-wide perturb-seq dataset (with scPRINT (Han et. al.’s heads) replaced with scPRINT (gwps’ heads): same scPRINT version but with attention heads selected using a subset of the genome-wide perturb-seq ground truth). Early Precision Ratio (EPR) and Area Under the Precision-Recall Curve (AUPRC) are provided here in one barplot, left to right. Source data are provided as a Source Data file.
