Improved estimation of cancer dependencies from large-scale RNAi screens using model-based normalization and data integration

James M. McFarland et al. Nature Communications 2018

Figure by Andrew Tang

RNAi Resources -


Improved gene dependency estimates for 712 cancer cell lines spanning 3 large-scale RNAi datasets


The availability of multiple datasets together comprising hundreds of genome-scale RNAi viability screens across a diverse range of cancer cell lines presents new opportunities for understanding cancer vulnerabilities. Integrated analyses of these data to assess differential dependency across genes and cell lines are challenging due to confounding factors such as batch effects and variable screen quality, as well as difficulty assessing gene dependency on an absolute scale.

To address these issues, we incorporated estimation of cell line screen quality parameters and hierarchical Bayesian inference into an analytical framework for analyzing RNAi screens (DEMETER2). We applied this model to individual large-scale datasets and show that it substantially improves estimates of gene dependency across a range of performance measures, including identification of gold-standard essential genes as well as agreement with CRISPR-Cas9-based viability screens. This model also allows us to effectively integrate information across three large RNAi screening datasets, providing a unified resource representing the most extensive compilation of cancer cell line genetic dependencies to date.

Read the full text

See the preprint on bioRxiv


The input files and model results from applying DEMETER2 to the following three large-scale RNAi datasets:

  1. 501 cancer cell lines from Project Achilles (Tsherniak et al. 2017)
  2. 397 cancer cell lines from Project DRIVE (McDonald et al. 2017)
  3. 76 breast cancer cell lines (Marcotte et al. 2016)


Clone or download our DEMETER2 GitHub repository to run DEMETER2 on RNAi screening data