Module 7: Long-Read Alignment Mechanics: Minimap2 Algorithmic Presets & Galaxy Workflows

1. Why Short-Read Aligners Fail on Long Reads

Traditional short-read aligners like BWA-MEM and Bowtie2 rely on exact Burrows-Wheeler Transform (BWT) seeds of $19-31\text{ bp}$. Because raw long reads contain indel error rates of $1-3\%$, exact short seeds are frequently disrupted, causing massive alignment failure and unacceptable computational overhead.

minimap2 solves this through an innovative three-stage heuristic:

1. Minimizer Sketching: Selects the lexicographically smallest $k$-mer within a sliding window of $w$ consecutive $k$-mers, dramatically reducing index memory while retaining robust anchors.
2. Colinear Chaining: Chains matching minimizers using a dynamic programming algorithm that permits non-linear gap penalties, identifying candidate alignment paths across tens of kilobases.
3. Banded Global Alignment: Performs full base-level Smith-Waterman alignment only within narrow bands around the chained anchors.

2. Crucial Minimap2 Presets for Long-Read Data

3. Production Alignment Execution

# Align ONT reads to the T2T-CHM13v2.0 or GRCh38 human reference genome
minimap2 -ax map-ont \
    -t 16 \
    --MD \
    -Y \
    -R "@RG\tID:SAMPLE1\tSM:PATIENT_01\tPL:ONT" \
    /references/GRCh38.fa \
    filtered_reads.fastq.gz | \
    samtools sort -@ 8 -m 2G -o aligned_sorted.bam -

# Index the coordinate-sorted BAM
samtools index -@ 8 aligned_sorted.bam

The --MD flag calculates mismatch and deletion strings essential for downstream variant callers, while -Y enables soft-clipping for supplementary alignments, preserving structural variant breakpoint signatures.