fastq-analysis_skill

Overview

This skill guides users through OmicVerse's alignment module to run end-to-end FASTQ-to-count workflows for bulk and single-cell RNA-seq. It covers SRA acquisition, FASTQ quality control, STAR alignment, gene quantification with featureCounts, and single-cell kallisto|bustools via kb-python. The skill emphasizes reproducible, parallel execution with automatic tool resolution and optional auto-installation.

How this skill works

The module exposes functions that wrap standard bioinformatics tools (prefetch, fasterq-dump, fastp, STAR, samtools, featureCounts, kb-python) and share a common CLI utility for resolving and auto-installing binaries. Workflows are composable: download SRA to FASTQ, run fastp for trimming/QC, align with STAR to produce coordinate-sorted BAMs, and quantify genes with featureCounts. For single-cell data, kb-python is used to build references and generate count matrices (h5ad/loom/MTX). Each function detects existing outputs and supports parallel jobs, retries, and streaming logs.

When to use it

• Download and validate SRA datasets and convert to FASTQ before analysis.
• Trim, filter, and generate per-sample QC reports with fastp.
• Run STAR alignment with automatic genome index building for bulk RNA-seq.
• Produce gene-level count matrices from BAMs using featureCounts for downstream differential analysis.
• Build kallisto|bustools references and quantify single-cell 10x or custom scRNA-seq libraries.

Best practices

• Run prefetch before fqdump for reliable SRA downloads and integrity checks.
• Keep auto_install=True in new environments to let the module resolve missing tools via conda/mamba.
• Use fastp outputs directly as STAR inputs by converting lists of result dicts to sample tuples.
• Enable auto_index=True for STAR to avoid manual index steps; provide genome fasta and GTF paths.
• Set overwrite=False to skip completed steps; use overwrite=True only to force re-runs when required.

Example use cases

• Bulk RNA-seq from SRA: prefetch -> fqdump -> fastp -> STAR -> featureCount -> pandas DataFrame counts.
• Local paired-end FASTQs: skip download, run fastp -> STAR -> featureCount for a lightweight local pipeline.
• Single-cell 10x v3: build kb-python reference with ref(), then count() to produce h5ad AnnData for downstream analysis.
• Parallel SRA processing: use fqdump and parallel_fastq_dump with jobs and threads to accelerate downloads and conversions.