Add comprehensive growth rate analysis and scoring script

[zorian15]

Growth Rate Analysis Script Plan with Data Export

Overview

Create scripts/score_growth_rates.py to evaluate growth rate predictions and export detailed data.

Script Structure & Outputs

1. Core Script: scripts/score_growth_rates.py

python scripts/score_growth_rates.py \
  --config configs/growth_rate_config.yaml \
  --build flu-simulated-150k-samples \
  --output-dir results/flu-simulated-150k-samples/growth_rate_analysis/

2. Output Structure:

results/{build}/growth_rate_analysis/
├── scores/
│   ├── growth_rate_scores.tsv          # Window-level metrics
│   └── variant_growth_rate_scores.tsv  # Variant-level metrics
└── growth_rates/
    ├── FGA/
    │   ├── growth_rates_north_2026-10-01.tsv
    │   ├── growth_rates_tropics_2026-10-01.tsv
    │   └── ...
    └── GARW/
        ├── growth_rates_north_2026-10-01.tsv
        └── ...

3. Growth Rate Data Files Content:

Each growth_rates_{location}_{pivot_date}.tsv contains:

• date, variant: Identifiers
• r_data, r_model: Core growth rate values
• median_r, r_lower_95, r_upper_95: Model estimates with CIs
• smoothed_frequency, raw_frequency: Variant frequencies
• smoothed_sequences, raw_sequences: Sequence counts
• case_counts, total_case_counts: Case data
• filtered: Boolean indicating if point passed filtering criteria

4. Key Functions:

def export_growth_rates_data(growth_rates_df, output_dir, model, location, pivot_date):
    # Save processed growth rates data to TSV files
    # Follows same naming convention as existing results/*/estimates/
    
def calculate_comprehensive_metrics(growth_rates_df):
    # Window-level and variant-level metrics using existing antigentools functions
    
def process_all_model_results(config):
    # Discover all rt_*.tsv files, process each window
    # Export both metrics and raw growth rate data

5. Configuration File: configs/growth_rate_config.yaml

models: ["FGA", "GARW"]
locations: ["north", "tropics", "south"]
filtering:
  min_sequence_count: 10
  min_variant_frequency: 0.01  
  min_total_sequences: 300
  epsilon: 1e-3
  min_segment_length: 3
spline_smoothing:
  factor: 1.0
  order: 3
export_options:
  save_growth_rates: true
  save_filtered_only: false  # Include all data points

6. Integration Benefits:

• Data export enables detailed downstream analysis
• Consistent format matches existing results structure
• Preserves provenance - all intermediate calculations saved
• Reproducible pipeline from raw model outputs to final metrics
• Easy visualization - growth rate files can be directly plotted

7. Usage in Analysis:

# Load exported growth rate data for specific window
gr_df = pd.read_csv("results/.../growth_rates/FGA/growth_rates_tropics_2028-10-01.tsv", sep="\t")

# Plot using existing functions
plot_growth_rate_dynamics(gr_df, location="tropics", ...)

This creates a comprehensive growth rate evaluation pipeline that both scores model performance AND preserves all the detailed data needed for publication figures and deep-dive analysis.