Extended Data Fig. 4: Pipelines used to identify negatively selected candidate circRNAs important for cell growth.

a, Analysis pipeline to identify negatively selected circRNA candidates by CDCscreen. Uniquely-mapped reads of each gRNA were normalized, and averaged normalized reads of two biological replicates were used for subsequent analyses. P value of negatively selected circRNA was computed by the permutation test of MAGeCK, and mean fold change of negatively-selected gRNAs targeting the same circRNAs between D30 and D1 treatments was obtained from normalized gRNA reads. Only expressed circRNAs with FPBcirc > 0 were calculated. Finally, circRNAs with CDCscreen score ≥ 2, ≥ 2 negatively-selected gRNAs with FC ≤ 0.667 were identified as circRNA candidates to promote cell proliferation. b, Rank of negatively selected candidate circRNAs by CDCscreen scores in HT29, 293FT and HeLa cells. CircRNAs with CDCscreen score ≥ 2 and with ≥ negatively-selected gRNAs and FC ≤ 0.667 were sub-grouped by red dashed line. Examples of candidate circRNAs were marked in each cell line. c, Each item in the table corresponds to each step of calculation of candidate circRNAs in CDCscreen pipeline shown in (a). CircFAM120A is shown as an example, and related analyses of all other circRNAs are listed in Supplementary Table 4. d, Analysis of gRNAs in FC ≤ 0.667 that identifies the same circRNA as a target gene. The number of circRNA candidates, the average total gRNAs and altered gRNAs in FC ≤ 0.667 (purple) in each cell line are shown. n = 67 circRNAs in HT29, n = 62 circRNAs in 293FT, n = 63 circRNAs in HeLa. Data are presented as means values ± s.d. e, False discovery rate (FDR) of circRNA candidates with CDCscreen score ≥ 2 is less than 0.1 for in vitro (Fig. 2b) and in vivo (Extended Data Fig. 7a) screens.