SkillAgentSearch skills...

gtars

Rust-backed Python library for fast genomic token arithmetic and BED processing. High-performance BED I/O, interval set ops (intersect, merge, complement, subtract), region tokenization against a universe, universe construction. Use for preprocessing large BED collections and ML token vocabularies.

Install / Use

npx skills add jaechang-hits/SciAgent-Skills --skill gtars

Installs into whichever agent you are using.

About this skill
📄

SKILL.md

Installable skill definition

Quality Score

91/100

Supported Platforms

Universal

Our assessment of gtars

gtars scores 91/100 on our quality scale, 1180th of 4,716 Development & Engineering skills we index (top 26%).

Its SKILL.md is 16 KB long, well organised into 59 sections with 18 code examples: a thorough specification that gives an agent plenty to work with.

It has 367 GitHub stars, a meaningful sign that others use it.

Substance
30/30
Structure
20/20
Description
15/15
Adoption
11/20
Freshness
15/15

Maintenance, license and trust

  • The repository was last updated about 3 months ago, so gtars is actively maintained.
  • Our last check on 2026-11-20 found the source still online.
  • No license is declared. By default that means all rights are reserved: you can read it, but reusing or redistributing it is not clearly permitted. Ask the author before building on it commercially.
  • Its trust signals score 88/100, with 1 caution from licensing, adoption, age or documentation. These come from repository metadata, not a code audit — read the skill file before letting an agent act on it.

Safety scan

No issues found

Our scan of the whole file found no instruction hijacking, hidden characters, credential access, data exfiltration or destructive commands. An AI review of the same text found nothing harmful.

AI review by kimi-k2.7-code on 2026-10-05. Automated pattern scan on 2026-10-05. It catches known dangerous patterns, not every risk — read a skill before letting an agent act on it.

gtars compared with similar skills

All 4 of these similar skills score higher than gtars; compare them before choosing.

SkillScoreStarsUpdatedFormat
gtars (this skill)by jaechang-hits913673mo agoSKILL.md
taste-skillby Leonxlnx100118.9k11d agoCLAUDE.md
Agent-Reachby Panniantong100113.3k12d agoCLAUDE.md
hyperframesby heygen-com10087.1ktodayCLAUDE.md
headroomby headroomlabs-ai10080.2ktodayCLAUDE.md

Frequently asked questions

How do I install gtars?
Run npx skills add jaechang-hits/SciAgent-Skills --skill gtars. The install tabs above show the steps for each supported agent.
Which AI agents does gtars work with?
It is written for Universal, as a SKILL.md file. Other agents that read the same format can often use it too.
Is gtars safe to use?
Our scan of the whole file found no instruction hijacking, hidden characters, credential access, data exfiltration or destructive commands. An AI review of the same text found nothing harmful. It declares no license and scores 88/100 on trust signals. Skills are instructions an agent will follow, so read the file before installing it and do not approve commands you do not understand.
Is gtars still maintained?
The repository was last updated about 3 months ago, so gtars is actively maintained.

name: "gtars" description: "Rust-backed Python library for fast genomic token arithmetic and BED processing. High-performance BED I/O, interval set ops (intersect, merge, complement, subtract), region tokenization against a universe, universe construction. Use for preprocessing large BED collections and ML token vocabularies." license: "MIT"

GTARS: Fast Genomic Token Arithmetic and BED File Processing

Overview

GTARS is a Python library with a Rust-backed core for high-performance genomic interval operations. It provides BED file I/O, set-theoretic interval operations (intersection, union, merge, complement, subtract), genomic region tokenization against a reference universe, and utilities for building consensus universe BED files. GTARS is designed for workflows that process hundreds to thousands of BED files efficiently, serving as a preprocessing engine for ML pipelines (including geniml) and general bioinformatics pipelines.

When to Use

  • Read and write large BED files efficiently, leveraging Rust-backed parsing for speed over pure Python alternatives
  • Compute genomic interval intersections, merges, complements, or subtracts between BED file pairs or sets
  • Tokenize a collection of genomic regions against a fixed universe vocabulary for ML input preparation
  • Build consensus universe BED files from a collection of sample BED files
  • Count overlap statistics between two BED files without launching bedtools processes
  • Preprocess ATAC-seq, ChIP-seq, or ENCODE peak files before feeding into geniml or other ML tools
  • For full BED/BAM/SAM reading with CIGAR-level detail, use pysam-genomic-files instead

Prerequisites

  • Python packages: gtars, numpy (optional, for array conversion)
  • Data requirements: BED format files (3+ columns: chr, start, end); optionally a universe BED file for tokenization
  • Environment: Python 3.8+; Rust toolchain not required (pre-built wheels available on PyPI)
pip install gtars

Quick Start

from gtars import GenomicIntervalSet

# Load a BED file and inspect
gis = GenomicIntervalSet("peaks.bed")
print(f"Loaded {len(gis)} intervals")
print(f"First interval: {gis[0]}")
# First interval: chr1:1000-2000

# Intersect two BED files
gis2 = GenomicIntervalSet("other_peaks.bed")
overlap = gis.intersect(gis2)
print(f"Overlapping intervals: {len(overlap)}")

Core API

Module 1: GenomicIntervalSet — BED File I/O

Primary data structure for loading, indexing, and writing genomic intervals.

from gtars import GenomicIntervalSet

# Load from file
gis = GenomicIntervalSet("peaks.bed")
print(f"Intervals loaded: {len(gis)}")
print(f"Chromosomes present: {gis.chromosomes}")

# Access by index
interval = gis[0]
print(f"chr={interval.chr}, start={interval.start}, end={interval.end}")

# Iterate all intervals
for iv in gis:
    width = iv.end - iv.start
    if width > 5000:
        print(f"Wide interval: {iv.chr}:{iv.start}-{iv.end} ({width} bp)")
        break
from gtars import GenomicIntervalSet, GenomicInterval

# Create from a list of GenomicInterval objects
intervals = [
    GenomicInterval("chr1", 100, 500),
    GenomicInterval("chr1", 600, 1200),
    GenomicInterval("chr2", 300, 800),
]
gis = GenomicIntervalSet(intervals)
print(f"Created GenomicIntervalSet with {len(gis)} intervals")

# Write to BED file
gis.to_bed("output_intervals.bed")
print("Saved output_intervals.bed")

Module 2: Interval Set Operations

Compute intersections, unions, merges, subtracts, and complements between BED sets.

from gtars import GenomicIntervalSet

peaks_a = GenomicIntervalSet("condition_A.bed")
peaks_b = GenomicIntervalSet("condition_B.bed")

# Intersection: intervals present in both sets
shared = peaks_a.intersect(peaks_b)
print(f"Shared intervals: {len(shared)}")

# Subtraction: intervals in A not overlapping B
a_only = peaks_a.subtract(peaks_b)
print(f"A-specific intervals: {len(a_only)}")

# Union (merge both sets, then merge overlapping)
combined = peaks_a.union(peaks_b)
print(f"Union intervals: {len(combined)}")
from gtars import GenomicIntervalSet

# Merge overlapping/adjacent intervals within a single set
gis = GenomicIntervalSet("fragmented_peaks.bed")
print(f"Before merge: {len(gis)} intervals")

merged = gis.merge()
print(f"After merge:  {len(merged)} intervals")

# Complement: genome-wide intervals NOT covered by peaks
# Requires chromosome sizes
chrom_sizes = {"chr1": 248956422, "chr2": 242193529, "chrX": 156040895}
complement = gis.complement(chrom_sizes)
print(f"Complement (uncovered) intervals: {len(complement)}")

Module 3: Tokenization

Convert genomic intervals to integer token IDs against a reference universe vocabulary.

from gtars import Tokenizer

# Initialize tokenizer with a universe BED file
tokenizer = Tokenizer("universe.bed")
print(f"Universe vocabulary size: {len(tokenizer)}")

# Tokenize a single BED file → list of token IDs
from gtars import GenomicIntervalSet
gis = GenomicIntervalSet("sample_peaks.bed")
tokens = toke…[redacted](gis)
print(f"Token IDs: {tokens[:10]} ...")
print(f"Total tokens: {len(tokens)}")
from gtars import Tokenizer, GenomicIntervalSet

tokenizer = Tokenizer("universe.bed")

# Tokenize and convert to numpy array for ML
import numpy as np
gis = GenomicIntervalSet("sample_peaks.bed")
tokens = toke…[redacted](gis)
token_array = np.array(tokens)
print(f"Token array shape: {token_array.shape}, dtype: {token_array.dtype}")

# Build a binary presence/absence vector over the full universe
vocab_size = len(tokenizer)
presence_vector = np.zeros(vocab_size, dtype=np.float32)
presence_vector[token_array] = 1.0
print(f"Presence vector shape: {presence_vector.shape}")
print(f"Fraction of universe covered: {presence_vector.mean():.4f}")

Module 4: Universe Building

Construct a consensus non-overlapping universe from a collection of BED files.

from gtars import UniverseBuilder

# Build universe from multiple BED files
bed_files = ["sample_1.bed", "sample_2.bed", "sample_3.bed", "sample_4.bed"]
builder = UniverseBuilder()
universe = builder.build(bed_files)

print(f"Universe regions: {len(universe)}")
universe.to_bed("consensus_universe.bed")
print("Saved consensus_universe.bed")
from gtars import UniverseBuilder

# Build with coverage threshold (region must appear in >= N% of samples)
bed_files = [f"chip_{i}.bed" for i in range(1, 21)]
builder = UniverseBuilder(fraction=0.25)   # present in >= 25% of samples
universe = builder.build(bed_files)
print(f"Universe (fraction>=0.25): {len(universe)} regions")

# Compare thresholds
for frac in [0.1, 0.25, 0.5]:
    b = UniverseBuilder(fraction=frac)
    u = b.build(bed_files)
    print(f"  fraction={frac}: {len(u)} regions")

Module 5: Interval Statistics and Utilities

Compute coverage, overlap counts, and basic statistics over BED files.

from gtars import GenomicIntervalSet

gis = GenomicIntervalSet("peaks.bed")

# Basic statistics
widths = [iv.end - iv.start for iv in gis]
import numpy as np
print(f"Interval count:       {len(gis)}")
print(f"Mean width (bp):      {np.mean(widths):.1f}")
print(f"Median width (bp):    {np.median(widths):.1f}")
print(f"Total coverage (bp):  {sum(widths):,}")

# Per-chromosome counts
from collections import Counter
chrom_counts = Counter(iv.chr for iv in gis)
for chrom, count in sorted(chrom_counts.items())[:5]:
    print(f"  {chrom}: {count} intervals")
from gtars import GenomicIntervalSet

# Overlap count matrix between two sets
peaks_a = GenomicIntervalSet("condition_A.bed")
peaks_b = GenomicIntervalSet("condition_B.bed")

# Count how many intervals in A overlap at least one interval in B
overlapping_a = peaks_a.intersect(peaks_b)
pct_overlap = len(overlapping_a) / len(peaks_a) * 100
print(f"A intervals overlapping B: {len(overlapping_a)}/{len(peaks_a)} ({pct_overlap:.1f}%)")

# Filter peaks by minimum width
min_width = 200
filtered = GenomicIntervalSet([iv for iv in peaks_a if (iv.end - iv.start) >= min_width])
print(f"Peaks >= {min_width} bp: {len(filtered)}/{len(peaks_a)}")
filtered.to_bed("filtered_peaks.bed")

Key Concepts

Tokenization and Universe Vocabulary

GTARS tokenization maps genomic regions to integer indices by finding the universe region that best overlaps each query interval. If a query interval does not overlap any universe region, it receives a special out-of-vocabulary (OOV) token. This vocabulary approach is analogous to NLP tokenization and enables treating genomic intervals as discrete tokens for transformer- and embedding-based models.

from gtars import Tokenizer, GenomicIntervalSet

tokenizer = Tokenizer("universe.bed")
# OOV token index is typically 0 or len(tokenizer)
print(f"OOV token ID: {tokenizer.unknown_token}")
gis = GenomicIntervalSet("test.bed")
tokens = toke…[redacted](gis)
oov_count = sum(1 for t in tokens if t == tokenizer.unknown_token)
print(f"OOV rate: {oov_count}/{len(tokens)} ({oov_count/len(tokens)*100:.1f}%)")

Common Workflows

Workflow 1: BED File Preprocessing Pipeline

Goal: Load raw ATAC-seq peaks, filter by width, merge overlaps, and tokenize for ML.

from gtars import GenomicIntervalSet, GenomicInterval, Tokenizer
import numpy as np

# Step 1: Load peaks
peaks = GenomicIntervalSet("atac_peaks_raw.bed")
print(f"Raw peaks: {len(peaks)}")

# Step 2: Filter by minimum width (remove very short noise peaks)
min_width = 150
filtered_intervals = [iv for iv in peaks if (iv.end - iv.start) >= min_width]
peaks_filtered = GenomicIntervalSet(filtered_intervals)
print(f"After width filter (>={min_width} bp): {len(peaks_filtered)}")

# Step 3: Merge overlapping peaks
peaks_merged = peaks_filtered.merge()
print(f"After merge: {len(peaks_merged)}")
peaks_merged.to_bed("atac_peaks_processed.bed")

# Step 4: Tokenize against universe
tokenizer = Tokenizer("universe.bed")
tokens = toke…[redacted](peaks_merged)
token_array = np.array(tokens)
print(f"Token array: {token_array.shape}, OOV rate: {(token_array == tokenizer.unknown_token).mean():.3f}")

# Step 5: Build presence vector for ML input
vocab_size = len(tokenizer)
presence = np.zeros(vocab_size, dtype=np.float32)
valid_tokens = token_array[token_array != tokenizer.unknown_token]
presence[valid_tokens] = 1.0
np.save("sample_presence_vector.npy", presence)
print(f"Saved presence vector: {presence.shape}, coverage={presence.mean():.4f}")

Workflow 2: Batch BED File Statistics

Goal: Compute overlap and coverage statistics across a collection of BED files.

from gtars import GenomicIntervalSet
from pathlib import Path
import pandas as pd
import numpy as np

bed_dir = Path("chip_peaks/")
bed_files = sorted(bed_dir.glob("*.bed"))
reference = GenomicIntervalSet("reference_regions.bed")

records = []
for bed_file in bed_files:
    gis = GenomicIntervalSet(str(bed_file))
    widths = [iv.end - iv.start for iv in gis]
    overlap = gis.intersect(reference)
    records.append({
        "sample":          bed_file.stem,
        "n_peaks":         len(gis),
        "mean_width_bp":   np.mean(widths),
        "total_coverage_bp": sum(widths),
        "overlap_with_ref": len(overlap),
        "pct_overlap":     len(overlap) / len(gis) * 100 if len(gis) > 0 else 0,
    })

df = pd.DataFrame(records)
print(df.to_string(index=False))
df.to_csv("bed_file_statistics.csv", index=False)
print(f"\nSaved bed_file_statistics.csv ({len(df)} samples)")

Workflow 3: Build Universe and Tokenize Corpus

Goal: From scratch — build a universe from a BED corpus and tokenize all files.

from gtars import UniverseBuilder, Tokenizer, GenomicIntervalSet
from pathlib import Path
import numpy as np

bed_dir = Path("atac_peaks/")
bed_files = [str(f) for f in sorted(bed_dir.glob("*.bed"))]
print(f"Building universe from {len(bed_files)} BED files...")

# Build universe
builder 

Truncated for display — read the full file on GitHub.

Related Skills

View on GitHub
GitHub Stars367
CategoryDevelopment
Updated2mo ago
Forks36

Languages

Python

Trust signals

88/100

From repository metadata: license, adoption, age and documentation. Not a code audit — see the Safety scan above for what the skill file itself contains.

1 medium