Initial EDA on 1,800 guide RNAs revealed:
- Guide RNAs containing TTTT repeats show mean efficiency of 0.47, compared to 0.78 for those without —> the strongest signal found so far
- GC content shows a mild positive correlation with efficiency (r = 0.07), consistent with the literature
- Class imbalance: ~84% of guides are high-efficiency —> this will require careful handling during model evaluation
- Which sequence features matter most? (Feature importance analysis in Week 3)
- Does position within the guide RNA matter —> e.g., are GC bases at the seed region more informative than at the 5' end?
- How does the model perform on guides targeting different gene families?
These questions will guide the next phases of the project.
Python 3 · pandas · scikit-learn · matplotlib · biopython