This md contains a bunch of scripts for generating data that didn't make into a paper.
conda activate clipnet
cd /home2/ayh8/clipnet/
mkdir -p ../predictions/k562/
python predict_ensemble.py \
../data/k562_data_folds/k562_sequence_nonintragenic_0.fna.gz \
../predictions/k562/k562_proseq_predictions_0.h5 \
--model_dir ../clipnet_k562/models/clipnet_k562_proseq \
--gpu 0
python calculate_performance_metrics.py \
../predictions/k562/k562_proseq_predictions_0.h5 \
../data/k562_data_folds/k562_proseq_0.npz \
../predictions/k562/k562_proseq_performance_metrics_0.h5conda activate clipnet
cd /home2/ayh8/clipnet_k562/
attr=profile
clipnet attribute \
-f ../data/k562/k562_centered.fa.gz \
-o ../attr/k562/k562_procap_deepshap_${attr}.npz \
-m models/clipnet_k562 \
-a $attr -c -vconda activate clipnet
cd /home2/ayh8/clipnet_k562/
attr=profile
clipnet attribute \
-f ../data/k562/k562_centered.fa.gz \
-o ../attr/k562/k562_proseq_deepshap_${attr}.npz \
-m models/clipnet_k562_proseq \
-a $attr -c -vconda activate bpnet
cd /home2/ayh8/attr/k562/
attr=profile
time modisco motifs \
-s attributions/k562_seqs_onehot.npz \
-a attributions/k562_proseq_deepshap_${attr}.npz \
-o attributions/k562_proseq_deepshap_${attr}_modisco.h5 \
-n 1000000 -l 50 -v
time modisco report \
-i attributions/k562_proseq_deepshap_${attr}_modisco.h5 \
-o attributions/k562_proseq_deepshap_${attr}_modisco/ \
-m ../data/JASPAR/JASPAR2024_CORE_vertebrates_non-redundant_pfms_meme.txt