Name		Name	Last commit message	Last commit date
Latest commit History 2 Commits
chr9_diagnostics		chr9_diagnostics
model		model
preprocessing		preprocessing
.gitignore		.gitignore
.pybiomart.sqlite		.pybiomart.sqlite
Deep_Learning_Day_Poster_26_Taner.pptx.jpg		Deep_Learning_Day_Poster_26_Taner.pptx.jpg
README.md		README.md
aggregate_ablation_sweep.py		aggregate_ablation_sweep.py
aggregate_seed_sweep.py		aggregate_seed_sweep.py
evaluate_LOGO.py		evaluate_LOGO.py
evaluate_LOGO_legacy_loop_bce.py		evaluate_LOGO_legacy_loop_bce.py
train.py		train.py
train_HVG.py		train_HVG.py

Repository files navigation

CS_1470_DL_Project

About

AD progresses gradually, yet most transcriptomic classifiers treat it as binary. I built a 1D CNN and trained it on hippocampal RNA-seq from 172 5xFAD/3xTgAD mice (4–18mo) with disease and age heads, achieving 0.974 AUROC (LOGO-CV). Since both models use Thy1 promoters, I ran ablations and multi-seed attribution to test for transgene artifacts.

Readme