Pysam - Genomic Alignments

Used for high-throughput sequencing pipelines. It allows efficient access to billions of DNA fragments aligned to a reference genome.

When to Use

• Processing next-generation sequencing (NGS) data.
• Analyzing genomic variants (SNPs, indels).
• Extracting reads from specific genomic regions.
• Building custom bioinformatics pipelines.
• Quality control of sequencing data.

Core Principles

Indexed Access

BAM files must be indexed (.bai) for efficient random access to genomic regions.

Coordinate System

Genomic coordinates are 0-based (Python-style) for positions, but 1-based for ranges in some contexts.

Read Attributes

Each read contains sequence, quality scores, alignment position, and flags.

Quick Reference

Standard Imports

import pysam

Basic Patterns

# 1. Open BAM file
samfile = pysam.AlignmentFile("aligned_reads.bam", "rb")

# 2. Iterate over reads in a specific genomic region
for read in samfile.fetch("chr1", 10000, 10100):
    print(f"Read: {read.query_name}, Quality: {read.mapping_quality}")
    print(f"Sequence: {read.query_sequence}")
    print(f"Position: {read.reference_start}")

# 3. Variant analysis (VCF)
vcf = pysam.VariantFile("mutations.vcf")
for rec in vcf.fetch("chr1", 10000, 10100):
    print(f"Pos: {rec.pos}, Ref: {rec.ref}, Alt: {rec.alts}")
    print(f"Genotype: {rec.samples['sample1']['GT']}")

# 4. Writing aligned reads
outfile = pysam.AlignmentFile("output.bam", "wb", template=samfile)
for read in samfile:
    if read.mapping_quality > 30:
        outfile.write(read)
outfile.close()

Critical Rules

✅ DO

• Always use indexed files - Create index with pysam.index("file.bam") for fast access.
• Check read flags - Use read.is_paired, read.is_unmapped to filter reads.
• Handle unmapped reads - Unmapped reads have reference_start = -1.
• Close files explicitly - Use context managers or .close() to avoid resource leaks.

❌ DON'T

• Don't iterate over entire BAM - Use fetch() with regions for efficiency.
• Don't ignore quality scores - Low-quality bases can cause false variants.
• Don't mix coordinate systems - Be consistent with 0-based vs 1-based indexing.

Advanced Patterns

Counting Reads per Gene

# Using a gene annotation file
genes = {}  # gene_name -> (chr, start, end)
for read in samfile.fetch():
    # Check if read overlaps any gene
    for gene, (chr, start, end) in genes.items():
        if read.reference_name == chr and start <= read.reference_start < end:
            genes[gene]['count'] += 1

Variant Filtering

# Filter variants by quality and depth
for rec in vcf.fetch():
    depth = rec.samples['sample1']['DP']
    qual = rec.qual
    if depth > 10 and qual > 20:
        # Process high-quality variant
        pass

Pysam provides the low-level access needed for genomic data processing, enabling researchers to work directly with the raw data of life itself.