preprocess.md

July 7, 2017 ยท View on GitHub

Basenji

Sequential regulatory activity predictions with deep convolutional neural networks.

Preprocess

bam_cov.py

Transform BAM alignments to a normalized BigWig (or HDF5-stored) coverage track.

ArgumentsTypeDescription
bam_fileBAM/CRAMAlignments file from which to extract coverage.
output_fileBigWig/HDF5Output coverage track stored as BigWig or HDF5, depending on the ".bw" suffix.
hdf5_fileHDF5Output HDF5 file with train_in/train_out, test_in/test_out and many other keys.

basenji_hdf5_single.py

Combine a set of coverage tracks stored as BigWig or HDF5 into a single file for training and testing, parallelizing over samples per-segment using multiprocessing on a single machine.

ArgumentsTypeDescription
fasta_fileFASTAFASTA file of chromosome sequences.
sample_wigs_fileText tableSample labels and paths to coverage files.
hdf5_fileHDF5Output HDF5 file with train_in/train_out, test_in/test_out and many other keys.

basenji_hdf5_cluster.py

Combine a set of coverage tracks stored as BigWig or HDF5 into a single file for training and testing, parallelizing over samples on our SLURM cluster.

ArgumentsTypeDescription
fasta_fileFASTAFASTA file of chromosome sequences.
sample_wigs_fileText tableSample labels and paths to coverage files.
hdf5_fileHDF5Output HDF5 file with train_in/train_out, test_in/test_out and many other keys.

basenji_hdf5_genes.py

Tile a set of genes and save the result in HDF5 for Basenji processing.

ArgumentsTypeDescription
fasta_fileFASTAFASTA file of chromosome sequences.
gtf_fileGTFGene annotations in gene transfer format.
hdf5_fileHDF5Output HDF5 file with gene sequences and descriptions.