poseigen_chisel is the biological-data processing package in the Poseigen ecosystem. It prepares genomic sequences, interval annotations, and signal tracks for machine-learning workflows.
Chisel keeps genomic structure explicit instead of flattening everything immediately, so sequences, signals, and annotations remain easy to inspect and trace.
Requires Python 3.12 or newer. Install from PyPI:
python -m pip install poseigen_chiselChisel's public helpers are re-exported from the poseigen_chisel package:
- Currents are Python lists of NumPy arrays, typically one array per chromosome or contig. A matching sequence-ID list (
BS_ids) identifies each array; signal arrays use positions as rows and signal channels as columns.Currents2Currentcombines multiple currents while preserving that per-contig organization. - Markers are pandas DataFrames with one row per genomic feature. Common columns include
subseq,loc,strand, andsize, with optional extra annotations. - Packs are fixed-size windows extracted around markers from a Current or sequence data. Signal profile packs typically have shape
(markers, positions, channels); sequence packs remain lists of sequences unless an encoding function such asSeqs2OHEis requested.
The package also provides helpers for working with these formats:
- Sequence processing includes
Seq2Int,Seqs2SigCurrent,Seqs2OHE, andSeqCurrent2Packfor encoding sequences and extracting sequence packs. - Markers and signal packs includes
Bed2Markersfor creating coordinate-labelled marker tables, plusSigCurrent2Packfor extracting signal profiles around markers. - Signal conversion includes
Bw2CurrentandBam2Currentfor converting BigWig tracks and BAM alignments into per-chromosome arrays. - Motif analysis includes
MEME2Motifs,PPM2PWM, andOHE_PWM_Scanfor importing motifs, preparing position-weight matrices, and scanning one-hot encoded sequences.
- Protein-DNA interactions
- Enhancer activity