GE

genomics-sv-detection

Analyze and classify structural variants from VCF data for bioinformatics workflows.

Install

mkdir -p .claude/skills/genomics-sv-detection && curl -L -o skill.zip "https://agentskills.codes/api/skills/download/11238" && unzip -o skill.zip -d .claude/skills/genomics-sv-detection && rm skill.zip

Installs to .claude/skills/genomics-sv-detection

Activation

This is the description your AI agent reads to decide when to run this skill — the better it matches your request, the more reliably it fires.

Load when summarising structural variants from an SV VCF (DEL / DUP / INV / TRA) — BND-notation parsing, size classification, per-type counts. Skip when working with small SNVs / indels (use `genomics-variant-calling`) or calling SVs from BAM (run Manta / Delly / Sniffles first).
280 chars✓ has a “when” triggerlonger than Claude Code's old 250-char listing cap (fine on current versions)
Intermediate

Key capabilities

  • →Parses SV-flavored VCF files
  • →Classifies structural variants by size
  • →Generates per-type counts for DEL/DUP/INV/TRA
  • →Resolves BND notation
  • →Produces summary CSV and Markdown reports

How it works

The tool reads INFO/SVTYPE fields from a VCF, calculates absolute SVLEN for size binning, and aggregates counts into a structured report.

Inputs & outputs

You give it
SV-flavored VCF file
You get back
CSV summary table and Markdown report

When to use genomics-sv-detection

  • →Generate summary table of structural variants
  • →Classify SVs by size and type
  • →Parse BND notation from VCF outputs

About this skill

genomics-sv-detection

When to use

Load this skill for the file-based analysis named in the description. The function library and CLI share the same calculations; no external aligner, assembler, caller or annotation service is started.

Use from a step

from skills._sdk.notebook import load_skill, read_input, write_output
library = load_skill("genomics-sv-detection")
data = read_input("input.vcf", reader=library.read_records)
result = library.analyze(data)
write_output(result, "tables/result.csv")
write_output(library.distribution_figure(result), "figures/distribution.png")

Run examples/example_step.py through the step runner for a small, hand-worked synthetic fixture. It asserts known summary values. The reader materializes the input in memory; use bounded FASTQ reads or pre-filter large genomic files before loading them.

API

<!-- api:begin generated from _api.py; regenerate with run.py api <skill dir> --write -->

read_records(path: str | Path) -> pd.DataFrame

Read records through read_input(path, reader=library.read_records).

:param path: Existing input file in the format documented under Inputs and outputs. :returns: Parsed records as a DataFrame. :raises ValueError: Input values or file structure cannot be parsed.

analyze(data: pd.DataFrame) -> pd.DataFrame

Compute sv-detection summaries and return a new table, leaving data unchanged.

:param data: Records containing chrom, sv_type, sv_len, filter, size_class, genotype.

:returns: Result table with diagnostics and summary in attrs['run_info']. :raises ValueError: Required columns are absent or records are empty or invalid.

run_info(data: pd.DataFrame, *, keep: bool=True) -> dict

Return the analysis diagnostics and summary.

:param data: Result returned by analyze. :param keep: Keep diagnostics by default; the CLI passes False. :returns: Independent diagnostics dictionary. :raises ValueError: analyze has not populated diagnostics.

distribution_figure(data: pd.DataFrame)

Plot sv_len values without writing files.

:param data: Result table containing sv_len. :returns: Matplotlib Figure. :raises ValueError: The value column is absent or the table is empty.

<!-- api:end -->

Methods and parameters

analyze returns a new DataFrame and leaves the input unchanged. run_info(result) returns the summary and method diagnostics. The CLI passes keep=False so diagnostics do not enter output tables. All calculations are deterministic; synthetic CLI demos retain seed 42.

Gotchas

  • read_records reads INFO/SVTYPE; breakend ALT notation without that field stays UNKNOWN. There is no mate-pair resolution or SV calling.
  • run_info()["summary"] uses absolute SVLEN and the legacy small (<1000), medium (<100000), and large size classes. Only the first sample genotype is read.

Inputs and outputs

Input files:

  • File types: .vcf

CLI output files:

  • tables/structural_variants.csv
  • report.md
  • result.json

The library writes no files. Steps use write_output; the CLI owns the listed artifacts. Public figure functions return matplotlib Figures and do not add new CLI outputs.

CLI

python skills/genomics/genomics-sv-detection/sv_detection.py --input input_file --output results/
python skills/genomics/genomics-sv-detection/sv_detection.py --demo --output /tmp/genomics_sv_detection_demo

See also

  • references/parameters.md
  • references/methodology.md
  • references/output_contract.md

Dependencies

numpy, pandas, matplotlib

When not to use it

  • →Calling SVs from BAM files
  • →Processing small SNVs or indels
  • →Resolving complex BND-to-TRA pairings without upstream processing

Prerequisites

pandasnumpy

Limitations

  • →Does not call SVs from BAMs
  • →Pure BND records without SVTYPE classify as UNKNOWN
  • →SVLEN is stored as absolute value

How it compares

It focuses specifically on summarizing existing SV callsets rather than performing the computationally intensive task of calling variants from raw alignments.

Compared to similar skills

genomics-sv-detection side by side with the closest alternatives in the catalog.

SkillInstallsUpdatedSafetyDifficulty
genomics-sv-detection (this skill)06moReviewIntermediate
llm-evaluation65moNo flagsAdvanced
evaluating-llms-harness310moReviewAdvanced
qutip411moReviewAdvanced

Try saying

Example prompts that trigger this skill in your AI assistant.

You might also like

llm-evaluation

wshobson

Implement comprehensive evaluation strategies for LLM applications using automated metrics, human feedback, and benchmarking. Use when testing LLM performance, measuring AI application quality, or establishing evaluation frameworks.

671

evaluating-llms-harness

davila7

Evaluates LLMs across 60+ academic benchmarks (MMLU, HumanEval, GSM8K, TruthfulQA, HellaSwag). Use when benchmarking model quality, comparing models, reporting academic results, or tracking training progress. Industry standard used by EleutherAI, HuggingFace, and major labs. Supports HuggingFace, vLLM, APIs.

337

qutip

davila7

Quantum mechanics simulations and analysis using QuTiP (Quantum Toolbox in Python). Use when working with quantum systems including: (1) quantum states (kets, bras, density matrices), (2) quantum operators and gates, (3) time evolution and dynamics (Schrödinger, master equations, Monte Carlo), (4) open quantum systems with dissipation, (5) quantum measurements and entanglement, (6) visualization (Bloch sphere, Wigner functions), (7) steady states and correlation functions, or (8) advanced methods (Floquet theory, HEOM, stochastic solvers). Handles both closed and open quantum systems across various domains including quantum optics, quantum computing, and condensed matter physics.

428

torchdrug

davila7

Graph-based drug discovery toolkit. Molecular property prediction (ADMET), protein modeling, knowledge graph reasoning, molecular generation, retrosynthesis, GNNs (GIN, GAT, SchNet), 40+ datasets, for PyTorch-based ML on molecules, proteins, and biomedical graphs.

326

string-database

davila7

Query STRING API for protein-protein interactions (59M proteins, 20B interactions). Network analysis, GO/KEGG enrichment, interaction discovery, 5000+ species, for systems biology.

217

transformer-lens-interpretability

davila7

Provides guidance for mechanistic interpretability research using TransformerLens to inspect and manipulate transformer internals via HookPoints and activation caching. Use when reverse-engineering model algorithms, studying attention patterns, or performing activation patching experiments.

215

Search skills

Search the agent skills registry