Integrate gpredomics v1.0.0 — regression, metadata variable selection, new algorithms

[eprifti]

Context

gpredomics v1.0.0 adds regression support, Pearson correlation, MCMC Gibbs, and 8 total algorithms. predomicsapp needs to integrate these changes.

Tasks

1. Metadata file upload and variable selection

• Allow uploading a metadata TSV file per dataset (alongside X and y)
• Parse metadata columns and show them in the UI
• Allow selecting a metadata column as the y variable (e.g., gene_count, age, BMI, Gram+_count)
• When a continuous variable is selected, auto-switch to regression mode (fit: spearman)
• When a categorical variable is selected (e.g., sex), use it as y for classification

2. Regression support in UI

• Add regression fit functions to parameter options: spearman, pearson, rmse, mutual_information
• When regression is selected, hide classification-specific params (threshold_ci, voting)
• Display regression metrics (Spearman/Pearson/RMSE) instead of AUC/sensitivity/specificity in results
• Update CSV report export for regression

3. MCMC Gibbs mode

• Add MCMC method selector: gibbs (default) / sbs
• Add p0 parameter (prior inclusion probability)
• Add n_chains parameter
• Update MCMC defaults: n_iter=1000, n_burn=500

4. Rebuild backend with gpredomics v1.0.0

• Update gpredomicspy for Vec y, pearson, MCMC params
• Rebuild Docker image
• Test all algorithms through the web UI