Fig. 4: Robustness of detection of OOR state with the ACR design.
From: Precise identification of cell states altered in disease using healthy single-cell references
a, Robustness to the size of the control cohort with the ACR and CR designs. Box plots of the AUPRC for simulations with an increasing number of donors in the control dataset (x axis), using the ACR (green) or CR design with query mapping (CR scArches, orange) or joint embedding (CR scVI, yellow). The results from simulations with five different OOR cell states, selected according to top mean TPR across designs in Fig. 3b and using five different samples of donors for each number of control donors and OOR state, are shown. In these simulations, five, seven or nine donors were used in the disease dataset (see Supplementary Fig. 1 for the full breakdown). b, Robustness to the size of the atlas dataset with the ACR and AR designs. Box plots of the AUPRC for simulations with an increasing number of studies in the atlas dataset, using the ACR (green) or AR (purple) design. The results from simulations with five different OOR cell states, selected according to top mean TPR across designs in Fig. 3b, are shown (see Supplementary Fig. 2 for a full breakdown).
