Extended Data Fig. 2: Additional figures showing proximity bias effects in rxrx3 and cpg0016.

a, Distribution plots of within-chromosome and between-chromosome cosine similarity for rxrx3 (left) and cpg0016 data (right). The within-chromosome distribution is shifted toward the positive, which was the initial indication that some bias was present. b, Scatterplot of gene-level one-sided Brunner-Munzel probabilities versus relative chromosome-arm position for three chromosome arms in the cpg0016 dataset. The value on the y-axis estimates the probability of an intra-chromosome-arm relationship involving a given gene having a higher cosine similarity than an inter-chromosome-arm relationship involving the same gene. c, Spearman correlations in plots similar to b across all chromosome arms for the cpg0016 data. The height of the bar for each arm agrees well with the degree of fading in diagonal blocks in Fig. 1c below the diagonal. Colors show Bonferroni-corrected p-values. d, Bulk RNA sequencing gene count depletion for cells treated with a ZNF394-targeting guide relative to untreated cells in 10-gene blocks across chromosome 7. Decreased expression is evident on the telomeric side of the cut site.