Skip to content

Latest commit

 

History

History
37 lines (24 loc) · 1.16 KB

README.md

File metadata and controls

37 lines (24 loc) · 1.16 KB

SVM_pipelines

Example use case on Kundaje lab cluster for SVM input generation:

/mnt/lab_data2/annashch/alzheimers_parkinsons/svm_pipeline/make_inputs

Execute scripts in make_inputs to generate input files for SVM model training.

Execute scripts in train_predict to train SVM models and get model predictions.

Execute scripts in score to get auPRC & other metrics of model performance.

Execute scripts in interpret to run gkmexplain interpretation.

Execute scripts in plot_interpretations to plot gkmexplain interpretations.

Execute util.gkmexplain_to_bigwig.py to generate a bigwig representation of gkmexplain scores

Note: SVM models for ENCODE tier 1 lines have been pre-trained and stored on oak at this location:

/oak/stanford/groups/akundaje/projects/chrombpnet/svm/dnase
/oak/stanford/groups/akundaje/projects/chrombpnet/svm/atac

Predictions for bQTL datasets ref & alt alleles are here:

/oak/stanford/groups/akundaje/projects/chrombpnet/svm/bQTL

Predictions genomewide for models trained on tier 1 DNASE data are here:

/oak/stanford/groups/akundaje/projects/chrombpnet/svm/genomewide_encode_tier1_lines_dnase