Setup

On first use, read setup.md for integration guidelines. Create ~/bioinformatics/ with user consent to store project context and preferences.

When to Use

User needs to analyze biological sequences, run genomic pipelines, or interpret sequencing data. Agent handles sequence alignment, variant calling, expression analysis, and format conversions.

Architecture

Memory lives in ~/bioinformatics/. See memory-template.md for structure.

CODEBLOCK0

Quick Reference

Core Rules

1. Verify Input Quality First

• - FASTQ: Run FastQC, check per-base quality, adapter content
• BAM: Verify sorted, indexed (samtools quickcheck)
• VCF: Validate format (bcftools view -h)

Bad input → garbage output. Always QC first.

2. Use Reference Genome Consistently

• - Human: GRCh38/hg38 (prefer) or GRCh37/hg19
• Mouse: GRCm39/mm39 or GRCm38/mm10
• Mixing references = invalid results

Store reference info in ~/bioinformatics/memory.md per project.

3. Preserve Raw Data

• - Work on copies
• Keep originals read-only
• Log every transformation step

4. Resource Awareness

• - Check file sizes before operations
• Use streaming when possible (samtools view | ...)
• Estimate memory needs (BWA: ~6GB for human genome)
• Warn before operations >10 minutes

5. Reproducibility

• - Log exact tool versions (samtools --version)
• Save command parameters
• Record input file checksums for critical analyses

Common Traps

• - Wrong chromosome naming — chr1 vs 1 causes silent failures. Check and convert with INLINECODE17
• Unsorted BAM — Most tools expect sorted input. Symptoms: errors or wrong results with no warning
• Index missing — BAM needs .bai, VCF needs .tbi. Commands fail cryptically without them
• Memory exhaustion — Large BAM operations kill the session. Stream or use --threads wisely
• Stale indices — After modifying BAM/VCF, regenerate index. Old index = corrupt reads
• 0-based vs 1-based coordinates — BED is 0-based, VCF/GFF is 1-based. Off-by-one bugs are common

File Formats Quick Reference

Essential Commands

Quality Control

Alignment

Variant Calling

Data Manipulation

Security & Privacy

Data access:

• - Only reads files user explicitly provides as input
• Writes outputs to directories user specifies
• Stores preferences in ~/bioinformatics/ (with consent)

Data that stays local:

• - All sequence data processed locally
• No external API calls for analysis
• Pipeline configs in ~/bioinformatics/

This skill does NOT:

• - Upload sequence data anywhere
• Access files without explicit user instruction
• Infer or collect data beyond explicit inputs
• Make network requests during analysis

Note: Installing tools (conda, brew) and downloading reference genomes requires internet access. These are user-initiated actions.

Related Skills

• - data-analysis — statistical interpretation
• INLINECODE31 — hypothesis testing
• INLINECODE32 — research methodology

Feedback

• - If useful: INLINECODE33
• Stay updated: INLINECODE34