Limits...
The Power Decoder Simulator for the Evaluation of Pooled shRNA Screen Performance.

Stombaugh J, Licon A, Strezoska Ž, Stahl J, Anderson SB, Banos M, van Brabant Smith A, Birmingham A, Vermeulen A - J Biomol Screen (2015)

Bottom Line: Using the negative binomial distribution, it models both the relative abundance of multiple shRNAs within a single screening replicate and the biological noise between replicates for each individual shRNA.We demonstrate that this simulator can successfully model the data from an actual laboratory experiment.The Power Decoder simulator is written in R and Python and is available for download under the GNU General Public License v3.0.

View Article: PubMed Central - PubMed

Affiliation: Dharmacon, part of GE Healthcare, Lafayette, CO, USA.

No MeSH data available.


Simulated and actual powers for both high- and low-noise screens. The power of each experiment with (A) two and (B) three replicates for actual (red) and simulated (blue) data. Error bars are the standard deviations of 30 normalizations for actual experiments and 900 simulations for simulated experiments. (C) The correlation between simulated and actual power for three-replicate experiments.
© Copyright Policy - open-access
Related In: Results  -  Collection

License 1 - License 2 - License 3
getmorefigures.php?uid=PMC4543901&req=5

fig4-1087057115576715: Simulated and actual powers for both high- and low-noise screens. The power of each experiment with (A) two and (B) three replicates for actual (red) and simulated (blue) data. Error bars are the standard deviations of 30 normalizations for actual experiments and 900 simulations for simulated experiments. (C) The correlation between simulated and actual power for three-replicate experiments.

Mentions: Knowing which shRNAs were enriched or depleted by a certain fold change, we are able to compute the exact power of each simulated screen after using DESeq19 to identify differentially expressed shRNAs. To determine whether these powers correlated well with actual powers calculated for the engineered screen, we applied the Monte Carlo method: for each screen, we used the Power Decoder simulator to generate 900 simulated screens by performing 30 normalizations on each experimental data set and then 30 simulations based on each normalization. Each simulated screen included three replicates. Plots of log2 fold change in counts versus average counts (commonly referred to as MA plots) for representative simulations of Screen 100_2x and Screen 500_2x are shown in Figure 3; they exhibit the same trends as the experimental data (Fig. 1), with lower noise and higher power as the shRNA fold representation increases, again demonstrating this approach’s ability to model experimental data. A comparison of the actual and modeled data for all experiments (Fig. 4; Table 1) shows that the simulated powers correlate closely with the true, experimentally determined powers. Because the simulated percentage powers are slightly larger than the true powers, with the overestimates having a mean of 6.16 and standard deviation of 4.12 presumably due to the slight differences between the modeled and true count distributions discussed above, they can usefully be treated as a realistic upper limit on possible sensitivity. Notably, the simulator predictions are equally reliable for Screen 500_1.5xPCR_100 and Screen 500_2xPCR_100, which have higher noise and lower power than the analogous screens with optimal PCR amplification; this demonstrates that the simulator is relevant even for screens that are very noisy.


The Power Decoder Simulator for the Evaluation of Pooled shRNA Screen Performance.

Stombaugh J, Licon A, Strezoska Ž, Stahl J, Anderson SB, Banos M, van Brabant Smith A, Birmingham A, Vermeulen A - J Biomol Screen (2015)

Simulated and actual powers for both high- and low-noise screens. The power of each experiment with (A) two and (B) three replicates for actual (red) and simulated (blue) data. Error bars are the standard deviations of 30 normalizations for actual experiments and 900 simulations for simulated experiments. (C) The correlation between simulated and actual power for three-replicate experiments.
© Copyright Policy - open-access
Related In: Results  -  Collection

License 1 - License 2 - License 3
Show All Figures
getmorefigures.php?uid=PMC4543901&req=5

fig4-1087057115576715: Simulated and actual powers for both high- and low-noise screens. The power of each experiment with (A) two and (B) three replicates for actual (red) and simulated (blue) data. Error bars are the standard deviations of 30 normalizations for actual experiments and 900 simulations for simulated experiments. (C) The correlation between simulated and actual power for three-replicate experiments.
Mentions: Knowing which shRNAs were enriched or depleted by a certain fold change, we are able to compute the exact power of each simulated screen after using DESeq19 to identify differentially expressed shRNAs. To determine whether these powers correlated well with actual powers calculated for the engineered screen, we applied the Monte Carlo method: for each screen, we used the Power Decoder simulator to generate 900 simulated screens by performing 30 normalizations on each experimental data set and then 30 simulations based on each normalization. Each simulated screen included three replicates. Plots of log2 fold change in counts versus average counts (commonly referred to as MA plots) for representative simulations of Screen 100_2x and Screen 500_2x are shown in Figure 3; they exhibit the same trends as the experimental data (Fig. 1), with lower noise and higher power as the shRNA fold representation increases, again demonstrating this approach’s ability to model experimental data. A comparison of the actual and modeled data for all experiments (Fig. 4; Table 1) shows that the simulated powers correlate closely with the true, experimentally determined powers. Because the simulated percentage powers are slightly larger than the true powers, with the overestimates having a mean of 6.16 and standard deviation of 4.12 presumably due to the slight differences between the modeled and true count distributions discussed above, they can usefully be treated as a realistic upper limit on possible sensitivity. Notably, the simulator predictions are equally reliable for Screen 500_1.5xPCR_100 and Screen 500_2xPCR_100, which have higher noise and lower power than the analogous screens with optimal PCR amplification; this demonstrates that the simulator is relevant even for screens that are very noisy.

Bottom Line: Using the negative binomial distribution, it models both the relative abundance of multiple shRNAs within a single screening replicate and the biological noise between replicates for each individual shRNA.We demonstrate that this simulator can successfully model the data from an actual laboratory experiment.The Power Decoder simulator is written in R and Python and is available for download under the GNU General Public License v3.0.

View Article: PubMed Central - PubMed

Affiliation: Dharmacon, part of GE Healthcare, Lafayette, CO, USA.

No MeSH data available.