Fig. 2: Performance of noise and cell type predictions on transcripts.

a Background noise prediction in the MERFISH cortex data using Bering for a specific field of view (FOV). The background noise annotated in the original paper of the data was shown on the left. b Distance distributions of molecules to their 16th nearest neighbor (x-axis) for the spots in (a) are shown. Fitted lines represent these distance distributions for spots predicted as background and foreground. c Jensen–Shannon divergence scores were computed to compare the distance distributions of background and foreground regions shown in (b) for individual FOVs (n = 15 biological replicates), as presented in the original paper and in the Bering prediction results. A one-sided t-test was performed, and the significance level is indicated at the top. The interpretation of box plots follows the same convention as in this figure (e). d Cell type prediction in the 10× Xenium data of Ductal Carcinoma In Situ (DCIS) using different transcript-level annotation methods, including TACCO and Bering with and without graph models (top). The zoomed-in visualization of a particular section of the tissue is presented below. e We evaluated the performance of TACCO and Bering quantitatively on cell type classification across FOVs (n = 15 biological replicates) using four key metrics: accuracy, macro F1 score, macro precision, and macro recall. Statistical significance was determined using one-sided Wilcoxon rank-sum tests, with p-values corrected using the False Discovery Rate (FDR) method (Benjamini/Hochberg). Boxplots represent the distribution of each metric, with the box spanning from the first to the third quartile and the median indicated by a horizontal line. Whiskers extend to the most extreme values within 1.5 times the interquartile range (IQR) from the quartiles. Statistical significance between models is indicated above each comparison: p < 0.05 (*), p < 0.01 (**), p < 0.001 (***), and p < 0.0001 (****). Corrected p-values are shown on the top if p < 0.05. Source data are provided as a Source Data file.