Supplementary Figure 2: Feature analysis of Cas12a guides.
(a) Schematic of exon-targeting hgRNA approach with CHyMErA. (b) HgRNA screening libraries were generated by performing two rounds of Golden Gate assembly. During the first step the synthesized 113-nt oligos containing both Cas9 and Cas12a guides were introduced into pLHCKO. During the second step, the spacer sequence between the two oligos was replaced with a hybrid scaffold consisting of the Cas9 tracrRNA followed by the Lb- or As-Cas12a direct repeat (DR). Schematic of Cas9 and Cas12a guide length, PAM sequence and double-stranded DNA cutting pattern is indicated at the bottom. (c) Dropout of Cas9 or Cas12a guides targeting essential genes in CGR8 cells. (d) Exonic Lb-Cas12a guides were grouped based on LFC cut-offs in the HAP1 and CGR8 optimization screens. Strongly depleting guides were use as positive, and neutral guides as negative cases. (e) Precision-recall (left panel) and receiver operating characteristic (right panel) curves of different machine-learning approaches for predicting Cas12a guide performance in HAP1 and CGR8 cells. CNN: convolutional neural networks; L1Logit: lasso regularized logistic regression. (f) Weblogo of filters learned by CHyMErA-Net in the convolutional layer. (g) Performance of exonic Lb-Cas12a hgRNAs grouped according to their PAM sequence. (n = 1,237 TTTA, 1,882 TTTC, 1,547 TTTG). Data derived from n = 3 independent technical replicates. Boxes show the interquartile range (IQR - 25th to 75th percentile), with the median indicated by a line. The whiskers extend to the quartile +/- 1.5 x IQR. (h) Enrichment analysis of active and inactive Lb-Cas12a guides based on chromatin accessibility from K562 cells.
