Figure 2.

Spatial reconstruction of mouse gastrulation cells using a single-cell-resolution ST reference

(A) Strategy for evaluating reconstruction performance. The training dataset consistently featured SeqFISH1 L1 data. We investigated two distinct scenarios: (1) evaluation using embryo 1 L2 data, sourced from the same embryo as the training data but originating from different samples; and (2) assessment using embryo 2 L2 and embryo 3 L2, derived from independent mouse embryos not utilized in the training phase.

(B) Evaluation of cell neighbor reconstruction. We quantified the performance of different methods by analyzing the average hit number within various k nearest neighbors across the three distinct mouse test datasets.

(C) Evaluation of neighborhood cell-type reconstruction. We computed the average Jessen-Shannon distance (JSD) for each cell type across the three mouse test datasets.