Figure 3

In silico predicted cleavage for the Drexel gRNAs predict 100% efficacy against patient-derived HIV-1 subtype B sequences from the held-out test cohort. (A) Depicts the predicted efficacy of each individual gRNA to cleave patient-derived HIV sequences. Each point represents the percentage of a patient’s vQS that are predicted to be cleaved by each gRNA for the 169 samples in the held-out cohort. The boxplots denote the quartiles, median and 95% confidence intervals. (B) Indicates the fraction 169 held-out samples with at least one gRNA predicted to cleave at least 70% of the sample. Column 1 represents the best performing gRNA as presented in (A). Column two represents the combination of the best and second-best gRNA combined. D-LTR-P4-227913 and D-LTR-P10-287206 denote combinations of the top-4 and top-10 gRNAs respectively. (C) Depicts the number of predicted samples cut from the test cohort with an increasing number of gRNAs across the 169 held-out samples. A threshold is indicated where the number of efficacious gRNAs was maximized. The grey box is expanded in (D). (D) An expanded view of the grey box from (C) demonstrates that D-LTR-P4-227913 and D-LTR-P10-287206 were predicted to specifically cleave more samples than by chance alone. (E) Indicates the percentage of times each gRNA was found to be in the top-10 best performing gRNAs in the validation cohort across 1000 iterations of randomly choosing 100 training samples and 169 testing samples.