Fast Python I/O for BigWig (continuous genome signal) and BigBed (interval annotation) files via libBigWig. Use for random-access signal queries at specific genomic coordinates (bw.values, bw.stats), computing per-region summary statistics (mean/max/coverage) over a BED file of regions, writing custom BigWig tracks from numpy arrays, and loading ChIP-seq/ATAC-seq/RNA-seq/methylation coverage tracks (e.g. produced by deeptools bamCoverage) into pandas/numpy for downstream analysis or ML feature extraction. Complements deeptools (which generates BigWig files) and chip-seq/atac-seq workflows.
Standardmäßig ist der Prompt ausgewählt, der zuerst die Quelle prüft. Sie können zu einem direkten Befehl wechseln oder eine lokale Kopie herunterladen.
Quelldateien prüfen
Lesen Sie SKILL.md und alle von SkillsMP angezeigten Begleitdateien, bevor Sie sich für eine Installation entscheiden.
Mit Codex oder Claude installieren Kopieren Sie diesen Prompt, fügen Sie ihn in Codex, Claude oder einen anderen Assistant ein und lassen Sie die Skill-Seite prüfen und installieren.
Ein direkter Befehl überspringt den Prüf-Prompt. Prüfen Sie die Quelle, bevor Sie ihn ausführen.
Fast Python I/O for BigWig (continuous genome signal) and BigBed (interval annotation) files via libBigWig. Use for random-access signal queries at specific genomic coordinates (bw.values, bw.stats), computing per-region summary statistics (mean/max/coverage) over a BED file of regions, writing custom BigWig tracks from numpy arrays, and loading ChIP-seq/ATAC-seq/RNA-seq/methylation coverage tracks (e.g. produced by deeptools bamCoverage) into pandas/numpy for downstream analysis or ML feature extraction. Complements deeptools (which generates BigWig files) and chip-seq/atac-seq workflows.
license
MIT
allowed-tools
["Read","Write","Edit","Bash"]
compatibility
Requires Python 3.9+ and pyBigWig 0.3.x (current 0.3.25). This is a C extension over libBigWig — bioconda ships prebuilt packages for linux-64, linux-aarch64, osx-64 and osx-arm64, while PyPI ships manylinux x86_64 wheels only, so `pip install pyBigWig` builds from source on macOS; that build needs libcurl and zlib headers (`curl-config` on PATH), which macOS supplies in the SDK.
metadata
{"version":"1.0","skill-author":"community"}
pyBigWig
Overview
BigWig files are the standard binary format for genome-wide continuous signal (read depth, fold enrichment, log2 ratio, per-CpG methylation) — the typical output of deeptools bamCoverage/bamCompare. BigBed files store interval annotations (peaks, genes) with optional scores in the same indexed binary family. pyBigWig is a thin, fast C-extension wrapper around libBigWig that gives you random access to either format: fetch signal for an arbitrary region in milliseconds without loading the whole (often multi-GB) file into memory.
Use it whenever an analysis needs to go beyond what deeptools' CLI already covers — pulling signal at custom coordinates for a model, correlating ChIP enrichment with expression at specific genes, or building a signal-aggregation plot with your own binning logic.
Installation
uv pip install pyBigWig
# or, to skip the libcurl/zlib build requirement entirely:
conda install -c bioconda -c conda-forge pybigwig
Import as import pyBigWig (mixed-case module name, despite the lowercase PyPI/conda package listing pybigwig).
When to Use
Extracting signal values at specific genomic coordinates for downstream ML feature vectors (e.g., ATAC-seq accessibility at a set of candidate enhancers).
Computing summary statistics (mean/max/coverage) over a custom BED of regions — differential peaks, gene bodies, TSS windows — without going back through the deeptools CLI.
Writing a BigWig track from a numpy array or custom per-base signal you've computed in Python.
Streaming through a BigWig/BigBed without materializing the whole file (chromosome-by-chromosome iteration).
Opening a BigWig directly from an HTTPS/FTP URL (e.g., a public ENCODE or UCSC track) without downloading it first.
Opening a File
import pyBigWig
bw = pyBigWig.open("sample.bw") # read mode (default)
bw_remote = pyBigWig.open("https://.../track.bw") # remote access works transparently
bwrite = pyBigWig.open("output.bw", "w") # write mode — can ONLY be written to, not queried
raises if the file doesn't exist, isn't a valid BigWig/BigBed, or is an unreachable URL — it never returns , so wrap it in / rather than testing the handle.
pyBigWig.open()
RuntimeError: Received an error during file opening!
# 0-based, half-open interval [start, end) — same convention as BED
values = bw.values("chr1", 1000, 2000) # list of floats, one per base; nan where no data
For large regions, prefer bw.stats(..., exact=False) (bin-level, uses precomputed zoom levels — fast) over pulling every base with bw.values and reducing in Python.
Bin-Level Statistics
# Mean signal in 100 equal-width bins across the region — fast, uses zoom levels
means = bw.stats("chr1", 0, 10000, type="mean", nBins=100)
# Single summary value for the whole region
mean_val = bw.stats("chr1", 0, 10000, type="mean")[0]
max_val = bw.stats("chr1", 0, 10000, type="max")[0]
coverage = bw.stats("chr1", 0, 10000, type="coverage")[0] # fraction of bases with data# Exact (base-pair-precise) instead of zoom-level-approximated — slower, use for small/critical regions
exact_mean = bw.stats("chr1", 0, 10000, type="mean", exact=True)[0]
type accepts "mean", "max", "min", "std", "coverage", or "sum". Bins with no data return None in the result list — handle that explicitly (see pitfalls).
Iterating Intervals
for start, end, value in bw.intervals("chr1"): # every stored interval on chr1
...
for start, end, value in bw.intervals("chr1", 1000, 5000): # restricted to a region
...
BigBed Entries
bb = pyBigWig.open("peaks.bb")
for start, end, rest in bb.entries("chr1", 0, 1000000):
# `rest` is a tab-separated string of any extra BED columns (name, score, strand, ...)
name, score, strand = rest.split("\t")[:3]
Region-Based Signal Extraction (BED → DataFrame)
The most common real-world pattern — mean signal per region from a BED file:
import pandas as pd
import pyBigWig
bw = pyBigWig.open("H3K27ac.bw")
regions = pd.read_csv("peaks.bed", sep="\t", header=None,
names=["chrom", "start", "end", "name"])
defmean_signal(row):
val = bw.stats(row.chrom, row.start, row.end, type="mean")[0]
return val if val isnotNoneelse0.0
regions["mean_signal"] = regions.apply(mean_signal, axis=1)
bw.close()
For thousands of regions, batching by chromosome and using bw.stats(..., nBins=1) per region is still the bottleneck-free approach — pyBigWig's stats call is C-level and fast per call, but calling it hundreds of thousands of times in a tight Python loop still dominates wall-clock time; vectorize with multiprocessing across chromosomes if this becomes a bottleneck.
Writing a BigWig File
import pyBigWig
bw = pyBigWig.open("custom_signal.bw", "w")
bw.addHeader([("chr1", 248956422), ("chr2", 242193529)]) # must be called before addEntries
bw.addEntries(
["chr1", "chr1", "chr1"],
[0, 100, 200],
ends=[100, 200, 300],
values=[1.5, 2.7, 0.3],
)
bw.close() # required — data isn't flushed to disk until close()
addEntries also accepts a single-chromosome + span/step form (bw.addEntries("chr1", 0, values=[...], span=10, step=10)) for fixed-interval tracks. addEntries accepts numpy arrays for chroms/starts/ends/values directly — no flag is needed — provided pyBigWig was compiled with numpy available; check with pyBigWig.numpy == 1. (There is no numpy keyword on open() or addEntries(). The numpy=True keyword is real but belongs to the read side, on values() and stats(), where it returns a numpy array instead of a Python list.)
Common Pitfalls
bw.stats() returns None for empty bins, not 0 or nan. Code that does arithmetic on the raw result list will crash or silently propagate None; explicitly substitute 0.0 (no signal) or float("nan") (missing data) depending on what's semantically correct for your analysis.
exact=True on huge regions is slow. It re-reads every base rather than using precomputed zoom-level summaries; only use it when you need base-pair precision (e.g., very small windows or footprinting), not for genome-wide scans.
Files opened "w" cannot be queried. A write-mode handle can only be written to (addHeader/addEntries) — open a second, separate read-mode handle if you need to verify what you just wrote.
Forgetting bw.close() after writing. BigWig writes are buffered; skipping close() (or crashing before it) leaves a truncated/invalid file.
Coordinate convention mismatch. pyBigWig follows BED's 0-based, half-open convention (values("chr1", 1000, 2000) covers positions 1000-1999). If you're cross-referencing 1-based coordinates from a VCF or GFF, convert (start - 1) before querying.
Remote file latency. Opening a BigWig over HTTPS/FTP works but each stats/values call may trigger a range request; for repeated heavy querying, download the file locally first.
Missing libcurl/zlib when building from source. If a prebuilt wheel isn't available for your platform, pip install pyBigWig fails with a linker error unless curl-config and zlib headers are installed — switch to the conda-forge/bioconda build instead of chasing system headers.
