Figure 5.

Batch effect correction for multimodal data with LinQ-View analysis

We applied LinQ-View on two CITE-seq datasets (PBMC 1K, Dataset 2; PBMC 10K, Dataset 3). Batch effects were removed by Seurat 3, and two modalities were integrated by LinQ-View. Heatmaps of ADT and RNA for cell populations identified from LinQ-View analysis with batch effect correction are shown on the bottom of (A) and (B). From left to right, four t-SNE embeddings were generated using RNA, ADT, joint, and joint dissimilarity matrix, respectively.

(A) A negative control without batch effect removal. Four cell populations affected by batch effects are indicated by green arrows and black dashed rectangles in the heatmaps.

(B) LinQ-View analysis with batch effect correction.