Fig. 2: Comparison of annotation results.
From: Knowledge-based inductive bias and domain adaptation for cell type annotation
a Comparison of accuracy and F1 on four datasets. There are n = 8 repeats with different random seeds. The error bars indicate mean ± s.d. b Inference time of various methods on the different data sizes, ranging from small to large. The comparison methods here do not include methods based on large models. c UMAP visualizations on BMMC (first row) and PBMC (second row). TOSICA is a SOTA lightweight model dedicated to RNA modality. Cellcano is a SOTA lightweight model dedicated to ATAC modality. d KIDA prediction results on 6 query batches (slices). Source data are provided in Supplementary Data 1.
