consenrich 0.1.0b0

Genome-wide extraction of reproducible continuous-valued signals hidden in noisy multisample functional genomics data

Consenrich

Consenrich is a sequential genome-wide state estimator for extraction of reproducible, spatially-resolved, epigenomic signals hidden in noisy multisample HTS data.

---

• Input:

  • $m \geq 1$ Sequence alignment files -t/--bam_files corresponding to each sample in a given HTS experiment
  • (Optional): $m_c = m$ control sample alignments, -c/--control_files, for each 'treatment' sample (e.g., ChIP-seq)

• Output: Real-valued 'consensus' epigenomic state estimates (BedGraph/BigWig) and uncertainty metrics.

---

• Robust, spatially informative consensus signal track representing multiple samples' epigenomic profiles $\implies$ Consenrich-extracted signal tracks can present additional insight for a variety of conventional analyses aiming to construct encompassing regulatory characterizations of sample groups (e.g., consensus peak calling)
• Consenrich is robust to scaling differences and models each sample's data and respective noise $\implies$ extract consensus signal tracks across HTS samples from different, related assays (e.g., ATAC-seq + DNase-seq, ChIP-seq + CUT-N-RUN)
• Consenrich can extract spectral features common to sample groups and is conducive to a wider range of signal processing-based analyses , e.g., targeted detection of structural/spatial patterns associated with specific regulatory properties/states.

Several technical features of Consenrich are discussed below.

Example Command-Line Use

• Run Consenrich on ten ATAC-seq samples in the current directory. Generate a BigWig signal track and inverse-variance-weighted residuals.

  consenrich --bam_files *.bam -g hg38 -o hg38_test_output.tsv --signal_bigwig demo_signal.bw --residual_bigwig demo_ivw_residuals.bw
  

---

• Use Consenrich for ChIP-seq enrichment analysis with treatment/control sample alignments (POL2RA, six donors' colon tissue samples). Generate separate BigWig output tracks for signal estimates and inverse-variance weighted residuals. Use fixed-width genomic intervals of 25bp:

  consenrich \
   --bam_files \
     ENCSR322JEO_POL2RA.bam \
     ENCSR472VBD_POL2RA.bam \
     ENCSR431EHE_POL2RA.bam \
     ENCSR724FCJ_POL2RA.bam \
     ENCSR974HQI_POL2RA.bam \
     ENCSR132XRW_POL2RA.bam \
   --control_files \
     ENCSR322JEO_CTRL.bam \
     ENCSR472VBD_CTRL.bam \
     ENCSR431EHE_CTRL.bam \
     ENCSR724FCJ_CTRL.bam \
     ENCSR974HQI_CTRL.bam \
     ENCSR132XRW_CTRL.bam \
   -g hg38 --step 25 \
   -o Consenrich_POL2RA.tsv \
   --signal_bigwig Consenrich_POL2RA_CTRL_Signal.bw \
   --residual_bigwig Consenrich_POL2RA_CTRL_IVW_Residuals.bw
  

Output

Download/Install

Consenrich can be easily downloaded and installed from source:

1. git clone https://github.com/nolan-h-hamilton/Consenrich.git
2. cd Consenrich
3. python setup.py sdist bdist_wheel
4. python -m pip install .
5. Check installation: consenrich --help

Consenrich is also available via PyPI/pip:

• pip install consenrich

If managing multiple Python environments, use python -m pip install consenrich. If lacking administrative privileges, running with flag --user may be necessary.

Technical Features

• Effectively models sample-and-region-varying noise to better integrate data across heterogeneous samples
• Balances biologically-informed a priori predictions with observed HTS data to determine final estimates
• Provides interpretable uncertainty quantification with respect to multiple model aspects
• Runs efficiently in linear time with respect to genome size.