scripts/7.explore.py
Performs preliminary exploration of the biological relationships in the final profiles: a UMAP embedding with controls highlighted, a heatmap, a mean average precision (mAP) test of whether replicates of a gene are distinguishable from the negative controls (using copairs), and similarity to positive controls.
Does not respect overwrite_files.
Inputs¶
| Input | Description |
|---|---|
profiles/{plate}_{data_set_name}_{gene|guide}_normalized_feature_selected_by{group}.csv.gz | From step 6 |
Config used: data_sets, control_genes, poscon_genes (experiment); feature_select.group, process_SBS.gene_col, process_SBS.barcode_col (defaults). See Configuration.
Outputs¶
All in data_exploration/.
| Output | Description |
|---|---|
{plate}_{data_set_name}_{gene|guide}_normalized_{group}_UMAP.png | UMAP of each gene/guide, colored by gene, with controls (hexagons) and positive controls (stars) marked. |
{plate}_{data_set_name}_{gene|guide}_normalized_{group}_heatmap.png | Heatmap of profiles by gene (gene-level) or gene and guide (guide-level) |
{plate}_{data_set_name}_guide_normalized_{group}_feature_selected_mAP.csv.gz | mAP and corrected p-value per gene, compared to negative controls |
{plate}_{data_set_name}_guide_normalized_{group}_feature_selected_mAP_scatterplot.png | mAP vs. -log10(p-value), positive controls marked |
{plate}_{data_set_name}_{gene|guide}_normalized_{group}_feature_selected_{gene}_cosine_similarity.csv | Cosine similarity of each profile to the positive control {gene} (only if poscon_genes is set) |
The mAP is calculated on the guide-level profiles of each plate.