-
Notifications
You must be signed in to change notification settings - Fork 12
Other Cell Lines
JasonLinjc edited this page Jul 14, 2026
·
1 revision
No CAGE labels exist for these four cells, so expression is RNA-only. Their enhancer encoders
train on pre-built activity CSVs, and the gene HDF5 is built from precomputed ABC links
(scripts/build_gene_h5_for_cell.py — no raw BAMs needed; contact = ABC average Hi-C).
Per cell (e.g. HepG2):
# 1) encoder
CELL=HepG2 sbatch slurm/train_seqencoder_12fold.slurm
# -> results/seqencoder/HepG2/checkpoints
# 2) gene HDF5 from precomputed ABC links
CELL=HepG2 sbatch slurm/build_gene_h5.slurm
# -> batch_output/HepG2/encoding/HepG2_samples.h5
# 3) expression (f3, RNA)
CELL=HepG2 EXPR_TYPE=RNA USE_PRM_SIGNAL=1 \
PRETRAINED_DIR=results/seqencoder/HepG2/checkpoints \
OUTPUT_DIR=EPInformer_models/HepG2_RNA sbatch slurm/train_epinformer_12fold.slurm
# 4) evaluate
python evaluate.py expression --pred_dir EPInformer_models/HepG2_RNAExpected RNA (pooled out-of-fold Pearson R): H1 0.781 · HepG2 0.845 · HUVEC 0.839 · NHEK 0.828.
These four cells' expression numbers are new — upstream EPInformer reports expression for K562/GM12878 only.