Reads Mapping

ChIP-seq / ATAC-seq Sequence Alignment (Bowtie2)

Overview

This skill performs core sequence alignment for ChIP-seq and ATAC-seq data starting from FASTQ files using Bowtie2. It is designed for autonomous execution with explicit user confirmation for biologically important parameters that must not be guessed.

Main steps include:

• Refer to the Inputs & Outputs section and create the output architecture in Step 0.
• Always ask the user for the reference genome/index to use. Never infer genome build by filename alone.
• Always ask the user for assay type if it cannot be confidently determined from file names or project context.
• Detect whether input data are paired-end or single-end.
• Group FASTQ files into samples using file naming conventions.
• Validate that the Bowtie2 index exists before running.
• Run alignment with Bowtie2.
• Convert SAM to BAM, then sort, index, and generate flagstat and idxstats reports.
• For ATAC-seq, preserve metadata and log that downstream duplicate handling and Tn5 shifting occur in later skills unless explicitly requested here.
• Write a per-sample parameter log file documenting all chosen options and their justification.

Inputs & Outputs

Inputs

Accepted FASTQ naming patterns include:

${sample}.fastq.gz
${sample}.fq.gz
${sample}_R1.fastq.gz
${sample}_R2.fastq.gz
${sample}_1.fastq.gz
${sample}_2.fastq.gz

Reference resources must be supplied by the user as one of the following:

/path/to/reference.fa
/path/to/bowtie2_index_prefix

Outputs

all_alignment/
  aligned_bam/
    ${sample}.sorted.bam
    ${sample}.sorted.bam.bai
  logs/
    ${sample}_alignment.log
    ${sample}_used_parameters.txt
  temp/

All outputs must be placed under ${proj_dir} returned in Step 0.

Required User Inputs

The agent must ask for the following when missing:

• Reference genome build (for example hg38, mm10, dm6)
• Reference asset path:
  • FASTA path for Bowtie2 index construction, or
  • existing Bowtie2 index prefix
• Assay type: ChIP-seq or ATAC-seq, if not obvious from file names or prior context
• Number of threads if the user has a compute preference; otherwise use a reproducible default
• Whether to keep intermediate SAM files; default is no

The agent must not guess:

• genome build
• reference annotation/version
• assay type when ambiguous
• single-end vs paired-end if file pairing is incomplete or inconsistent

Decision Logic

Assay Type

Use file and sample names only for tentative classification:

• Names containing ATAC, OmniATAC, scATAC → likely ATAC-seq
• Names containing TF or histone mark identifiers such as CTCF, MYC, H3K27ac, H3K4me3, H3K27me3 → likely ChIP-seq

If naming is ambiguous, ask the user.

Library Layout

Use FASTQ grouping rules:

• If files appear as ${sample}_R1 and ${sample}_R2 or ${sample}_1 and ${sample}_2 → paired-end
• If only one FASTQ exists for a sample → single-end
• If an R1 file exists without its mate → stop and ask the user to resolve missing mates

Threads

• Default to 8 threads unless the user specifies otherwise

Step-by-Step Workflow

Step 0: Initialize Project

Create a task directory for alignment outputs.

Suggested call:

• mcp__project-init-tools__project_init

with:

• sample: all
• task: alignment
• genome: provided by user

The tool will return ${proj_dir}. Use it for all output placement.

If a project-init MCP tool is not available in the runtime, create this directory structure manually:

all_alignment/
  aligned_bam/
  qc/
  logs/
  temp/

Set ${proj_dir} to all_alignment.

Step 1: Detect and Group FASTQ Files

Call:

• mcp__bowtie2-tools__detect_fastq_samples

with:

• input_dir: directory containing FASTQ files

The tool will:

• find supported FASTQ files
• group them into samples
• detect single-end vs paired-end layout
• report any missing mate files
• return a machine-readable sample summary

Rules:

• Prefer compressed FASTQ (*.fastq.gz, *.fq.gz)
• Use consistent sample grouping
• Fail if file naming is inconsistent or ambiguous

Step 2: Validate or Build Bowtie2 Reference Assets

If the user supplied a FASTA, call:

• mcp__bowtie2-tools__build_bowtie2_index

with:

• reference_fasta: user-provided FASTA path
• index_prefix: desired Bowtie2 index prefix

If the user supplied an existing Bowtie2 prefix, validate it before alignment by calling:

• mcp__bowtie2-tools__validate_bowtie2_index

with:

• index_prefix: user-provided index prefix

Stop and ask the user to correct the path if validation fails.

Step 3: Align Each Sample

For each detected sample, call:

• mcp__bowtie2-tools__run_bowtie2_alignment

with:

• sample_name: sample identifier
• fastq_r1: path to R1 FASTQ or single-end FASTQ
• fastq_r2: path to R2 FASTQ for paired-end data, otherwise omit
• assay_type: chipseq or atacseq
• index_prefix: validated Bowtie2 index prefix
• out_dir: ${proj_dir}/aligned_bam
• log_dir: ${proj_dir}/logs
• threads: user-specified or default 8
• keep_sam: false by default

Tool behavior:

• run Bowtie2
• write an alignment log
• convert to BAM
• sort BAM
• index BAM
• remove intermediate SAM unless keep_sam=true

Expected output:

${proj_dir}/aligned_bam/${sample}.sorted.bam
${proj_dir}/aligned_bam/${sample}.sorted.bam.bai
${proj_dir}/logs/${sample}_alignment.log

Notes:

• For ATAC-seq, this skill performs alignment only. Duplicate marking/removal, mitochondrial filtering, proper-pair filtering, and Tn5 shifting belong to downstream preprocessing or peak-calling skills unless explicitly requested in another skill.
• Do not remove duplicates here unless the user specifically requests an alternate alignment workflow.

Step 4: Write Parameter Log

For each sample, the agent must write:

${proj_dir}/logs/${sample}_used_parameters.txt

Example content:

Sample: ATAC_rep1
Assay type: ATAC-seq
Library layout: paired-end
Aligner: bowtie2
Reference genome build: hg38
Reference index: /refs/hg38/bowtie2/hg38
Threads: 8
Intermediate SAM kept: no

Reasoning:
- Sample name contains ATAC, so assay classified as ATAC-seq
- Paired FASTQ mates were detected automatically
- User provided hg38 Bowtie2 index
- Alignment-only workflow selected; duplicate handling and Tn5 shifting deferred to downstream preprocessing/peak-calling

Failure Handling

Stop execution and ask the user for correction if any of the following occurs:

• no FASTQ files found
• inconsistent file naming prevents sample grouping
• paired-end mate missing
• reference FASTA or index prefix missing
• bowtie2 or bowtie2-build executable not found in PATH
• samtools not found in PATH
• alignment command returns non-zero exit status
• sorted BAM or BAM index is not created

Do not continue to downstream QC if alignment fails.

Exact MCP Tool Calls Required

1. mcp__bowtie2-tools__detect_fastq_samples
2. mcp__bowtie2-tools__build_bowtie2_index (only when user provides FASTA)
3. mcp__bowtie2-tools__validate_bowtie2_index
4. mcp__bowtie2-tools__run_bowtie2_alignment

When the Agent Must Ask the User

The agent must ask before execution when any of the following are missing or ambiguous:

• genome build
• reference FASTA or Bowtie2 index prefix
• assay type when not inferable from context
• whether incomplete FASTQ pairs should be excluded or fixed
• thread count if a project-specific compute policy exists

The agent must not invent these values.