Fig 1.
Overall workflow diagram.
Fig 2.
(A) Before (B) After correction.
Fig 3.
Distribution of cell counts across distinct cell types for the first fold of integrated pancreatic dataset.
Table 1.
Evaluation metrics values between the synthetic and test data for PBMC3K.
Table 2.
Evaluation metrics values between the synthetic and test data for PBMC68K.
Table 3.
Evaluation metrics values between the balanced train(original and synthetic) and test data for integrated pancreatic and HCA-BM10K datasets.
Fig 4.
Distribution of original training, generated (synthetic) and test cells in 2-dimensional space where data generation is done by(A) MOE-FB, (B) MAF-FB, (C) CTGAN, (D) TVAE, (E) GC.
Fig 5.
Example of mutually exclusive DEGs found in literature.
Fig 6.
Total number of significant interactions between cell types for (A) Test set, (B) MOE-FB, (C) MAF-FB, (D) TVAE, (E) CTGAN, (F) GC.
Table 4.
Comparison of generative models based on average expression levels of cell–cell interactions.
Fig 7.
Ligand-receptor pairs representing the cell-cell interactions for the test set of integrated pancreatic dataset.
Table 5.
Datasets.