Fig. 3: The degree of network convergence is influenced by the functional similarity of target perturbations.

A Defining convergence and calculating convergent network strength. Here, we define convergence as the independent development of transcriptomic similarities between separate gene perturbations that move towards union or uniformity of biological function. B–G Principal component analysis (PCA) of the convergence scores, the three Gene Ontology scores (Molecular Function, M.F.; Biological Process, B.P.; Cellular Component, C.C.), brain expression correlation (B.E.C), and sample size across all resolved networks in both the pooled and arrayed assays revealed that some functional scores have similar influence on variance as convergence (SI Fig. 19). B, E Distribution of the degree of convergence (x axis) of networks across all possible combinations of 2–8 (y axis; number of sets tested within each set) target perturbations from the single-cell pooled experiment (B) and arrayed experiment (E) across all possible combinations of 2–14 target perturbations from the arrayed experiment show an influence of sample size on the ability to resolve a network. Data are represented as median values, lower and upper hinges correspond to the 1st and 3rd quartiles, upper and lower whiskers represent the largest values within 1.5*IQR (inter-quartile range) from the first or third quartile. Each point represents convergence based on biclustering between 2–8 unique combinations of CRISPR perturbations. B N = 4 replicates per condition (2 x donors, 2 x independent replicates per donor). E N = 2 biological replicates, 10 gRNA replicates (SCZ1); N = 2 biological replicates, 2 technical replicates (sequencing batches) and 3 gRNA replicates (SCZ2). C, F For both the pooled (C) and the arrayed (F) experiment, PCs 1 (x axis) and 2 (y axis) explain ~62% of the variance between networks. PC loadings demonstrate the influence of each variable on the variance between networks; within the first two PCs, the influence of brain expression correlation (B.E.C) and proportion of signaling genes perturbed (S.P) on PCs 1 and 2 on variance explained is more strongly related to convergence degree compared to other functional scores. Since degree of convergence is influenced by number of eGenes perturbed, we ran PCA analysis within networks of the same set size and found that the pattern of influence of signaling proportion and brain expression correlation is maintained when convergence is ranked within set size shown in SI Fig. 20. D, G This corresponds to an overall significant positive correlation between network convergence degree, signaling/synaptic proportion of perturbed genes in a set, and brain expression correlation between genes in a set (Bonferroni adjusted p value of Pearson’s correlations: *<0.05, **<0.01, ***<0.001. Created with BioRender.com.