Implement methods for phenotype calculation and statistical analysis

[abearab]

preprocessing

• Implement function to add pseudocounts

phenoscore

• Split scripts for phenotype score calculation and statistical analysis #25
• Add DESeq2 test as an option for differential analysis #71
• Formulate the "combined score" based on the given phenotypic scores and p-values. #80
• enable extracting genetable_collapsed by leveraging guides targeting multiple transcripts of same gene #85
• Z-score normalization relative to negative control distribution #17
• Write function to run robust ranking aggregation (RRA) algorithm #87
• Find and filter out low counts in a pair-wise comparison #16
• Compare sequencing count normalization across all samples or in pair-wise sample within comparisons
• Implement methods for nominating synthetic lethal interactions #72
• Unexpected fold change direction in phenoscore analysis of recombinations #73
• Gold standard AUC/ROC hit comparison across pipelines #4

phenostats

• Benjamini-Hochberg FDR calculation #11
• Reimplement CRISPhieRmix approch #34
• Implement simple stats.ttest_ind test #82
• Implement empirical p-value formula by comparing with non-targeting distribution #83
• Implement Kolmorogov-Smirnov test
• Implement Mann–Whitney U test

---

related papers / tools / scripts:

• Reproducibility metrics for CRISPR screens | bioRxiv | Cell Systems
• Fudging the volcano-plot without dredging the data
  https://www.nature.com/articles/s41467-024-45834-7
• CurveCurator: a recalibrated F-statistic to assess, classify, and explore significance of dose–response curves
  https://www.nature.com/articles/s41467-023-43696-z
  • https://github.com/kusterlab/curve_curator

---

root – GilbertLabUCSF/ScreenPro2#19

[abearab]

Nick

Why average among replicates instead of including that technical variance in the model?

Abe

How would you do that? Any examples?

Nick

Use a model that allows for intra-sample variation (e.g., a linear mixed model or other hierarchical model).