Uncovering hidden cancer self-dependencies through analysis of shRNA-level dependency scores

Area of Science:

• Computational biology
• Cancer genomics
• Functional genomics

Background:

• Large-scale short hairpin RNA (shRNA) screens are crucial for identifying cancer dependencies in human cancer cell lines.
• Off-target effects of shRNA reagents present a significant challenge in analyzing screen data.
• Current computational methods often rely on aggregated gene-level scores, potentially masking true dependencies.

Purpose of the Study:

• To develop a novel computational method, NBDep, for identifying cancer dependencies.
• To directly analyze shRNA-level dependency scores, bypassing the limitations of gene-level aggregation.
• To improve the accuracy and robustness of cancer dependency discovery in large-scale screens.

Main Methods:

• Developed the NBDep algorithm to analyze shRNA-level dependency scores directly.
• Implemented batch effect removal and selection of concordant shRNAs for each gene.
• Utilized negative binomial random effects models to assess gene-viability associations across cell lines.

Main Results:

• NBDep identified more known and putative cancer genes than alternative gene-level approaches in pan-cancer and cancer-specific analyses.
• The method was applied to shRNA dependency scores from Project DRIVE, covering 26 cancer types.
• Simulation studies demonstrated that NBDep effectively controls type-I error and outperforms gene-level score-based tests.

Conclusions:

• NBDep offers a more accurate and robust approach to discovering cancer dependencies from shRNA screen data.
• Directly analyzing shRNA-level scores mitigates off-target effects, enhancing the reliability of identified cancer dependencies.
• The NBDep method shows promise for advancing cancer research and therapeutic target identification.