Fig. 4From: Seed-effect modeling improves the consistency of genome-wide loss-of-function screens and identifies synthetic lethal vulnerabilities in cancer cellsReproducibility of the genome-wide shRNA screens after accounting for seed sequence properties. Two seed sequence properties were investigated: seed pairing stability (SPS) and target site abundance (TA). Rank correlation (ρ) over the 17 high data quality cell lines for shES of shRNAs a with strong (S) or weak (W) SPS, b with low (L) or high (H) TA, or c combined. Correlation for shES of shRNAs with position 12–18 heptamers after the same categorization is also shown as a reference. shRNAs with higher off-target seed sequence proficiency (i.e., strong SPS and low TA) show decreased consistency between the two studies. Asterisks denote statistically significant differences in correlation (p < 0.05, paired t-test). Strong SPS was defined as the top 10% percentile (SPS < −9.82), and weak SPS as the bottom 10% percentile (SPS > −5.16). Low TA >3.72 and high TA <2.89 were defined similarly, as shown at the top of each panelBack to article page