Fig 3 -

scAEGAN shows robust performance, while integrating datasets from the same platforms, a) scaAEGAN outperforms other methods for integrating a real scRNA-seq SMARTseq dataset from two mouse strains (Old and Young). A2B and B2A are the integration results of the Old and Young mice datasets with the latent spaces of input dataset from these two mouse strains, b) shows the heatmap of the cluster similarity of latent space of old mice dataset (AE_Old) with the translated domain B2A and likewise for young mice dataset (AE_Young Vs A2B) and c) shows the ARI and JI values for Old and Young mice dataset, depicting the outperformance of scAEGAN as compared to other methods. d) ARI and JI values for both simulated datasets A and B, depicting the outperformance of scAEGAN with AE-concatenated and Seurat, e) scAEGAN performs better, even the certain percentage of cells are removed from two datasets, f) scAEGAN shows robust performance, when there is an imbalance of cell types in two datasets (denoted by No overlap with A (cluster present in dataset A only and not in dataset B), 10_A(10% of cells removed in dataset A from cluster 6_1) and likewise for 50_A and 100_A respectively.