# Research pipeline index

The public source subset is grouped by purpose; chronological numbers preserve the frozen execution lineage. Exact scripts containing donor/cell identifiers are withheld and described in `protocols/REDACTED_FROZEN_STAGES.md`.

## Environment and upstream inventory

- `01_system_preflight.py`
- `03_inventory_manifest.py`
- `04_c2s_smoke_test.py`

## Metadata and taxonomy

- `05_audit_small_metadata.py`
- `06_preview_cell_metadata.py`
- `07_download_full_metadata_and_taxonomy.py`
- `08_build_mtg_annotated_metadata.py`
- `09_audit_mtg_taxonomy_and_qc.py`

## Frozen donor design and MTG matrix

- `10_generate_donor_split_candidates.py`
- `11_freeze_donor_split.py`
- `12_download_mtg_raw_matrix.py`
- `13_inspect_mtg_raw_matrix_v3.py`
- `14_build_primary_cell_index.py`
- `15_create_sampling_manifests.py`

## Ranked-gene and C2S representation

- `18_rebuild_ranked_gene_records_c2s_aligned.py`
- `19_full_development_token_audit.py`
- `20_create_c2s_base_pilot_manifest.py`
- `21_run_c2s_base_model_pilot.py`
- `22_compare_c2s_base_pilots.py`
- `23_create_score_vector_manifests.py`
- `24_extract_c2s_exact_score_vectors.py`
- `25_fit_c2s_score_vector_classifier.py`
- `26_audit_score_vector_signal.py`
- `27_extract_c2s_frozen_embeddings_v2.py`
- `28_fit_c2s_embedding_probe.py`
- `29_audit_embedding_probe_signal.py`

## Classical and raw-expression baselines

- `30_build_classical_rank_features.py`
- `31_fit_classical_rank_baselines.py`
- `32_compare_development_baselines.py`
- `33_build_raw_hvg_features.py`
- `34_fit_raw_hvg_baseline.py`
- `35_create_canonical_marker_masked_manifests.py`
- `36_fit_canonical_marker_masked_baselines.py`
- `37_summarize_in_region_development.py`

## Assay transfer

- `38_build_multiome_transfer_manifests.py`
- `39_build_multiome_ranked_gene_records.py`
- `40_build_multiome_transfer_features.py`
- `41_evaluate_multiome_transfer.py`

## Region transfer

- `42_audit_region_transfer_candidates.py`
- `43_download_mec_raw_matrix.py`
- `44_inspect_and_build_mec_transfer_manifests.py`
- `45_build_mec_ranked_gene_records_v2.py`
- `46_build_mec_transfer_features.py`
- `47_evaluate_mec_transfer.py`

## Donor-level AD severity

- `48_audit_ad_progression_feasibility.py`
- `50_build_ad_severity_donor_features.py`
- `51_fit_ad_severity_training_baselines_v1_1.py`
- `52_evaluate_ad_severity_validation.py`

## Deliberately withheld exact sources

- frozen MEC one-cell compatibility exclusion wrapper: contains an exact cell and donor identifier;
- frozen AD feature-cohort bias-audit script: contains exact donor identifiers;
- frozen donor split configuration: contains donor-level assignments;
- all prediction, embedding, matrix, and joblib artifacts.