Viewing Alignments and SNPs

Last updated on 2026-09-03 | Edit this page

Estimated time: 45 minutes

Overview

Questions

  • How do I inspect individual sequencing reads that support a candidate variant?
  • How can I tell a real SNP apart from a technical artifact?
  • What does a BAM alignment record actually contain?

Objectives

  • Load aligned sequencing reads from a BAM file alongside a BED file of candidate variant sites.
  • Read mismatched bases, coverage, and allele counts from an alignment track.
  • Sort and color aligned reads to help distinguish real variants from artifacts.
  • Use BLAT to check whether a suspicious signal is a genuine mapping or a multi-mapping artifact.

Background: paired-end sequencing and the BAM format

The alignment file you will load in this episode comes from paired-end Illumina sequencing: each DNA fragment is read from both ends, producing two reads (Read 1 and Read 2) separated by an unsequenced gap. The distance between the two reads’ outer edges is the insert size.

Diagram of single-end and paired-end sequencing reads aligned to a reference sequence
Single-end vs. paired-end reads, and the relationship between read length, inner distance, and insert size.

Aligned reads are stored in SAM/BAM format (Sequence Alignment/Binary Alignment Map — BAM is just a compressed, indexed binary encoding of SAM). Each alignment record includes the read name, its mapping position, a mapping quality score, a CIGAR string describing matches/insertions/ deletions along the alignment, information about its paired mate, and the read’s sequence and per-base quality scores.

Annotated example of SAM format fields
An example SAM file: a header describing the reference, followed by one alignment record per line, with columns for read name, position, mapping quality, CIGAR string, mate info, and sequence.

You do not need to read SAM/BAM records by hand to use IGV — this is what IGV’s alignment track visualizes for you — but recognizing these terms (insert size, mapping quality, CIGAR) will help you make sense of the popups and sort/color options used throughout this episode.

Reference genome

Make sure the reference genome is still set to the one you loaded in the previous episode: select chr1.fasta from the genome drop-down menu if needed.

Load the alignment and candidate SNP data

Select File > Load from File…, navigate to igvData/snps/, and open:

  • snp_calls.bed — a BED file marking two candidate SNP loci, and
  • NA12878.SLX.sample.bam — a small sample alignment file.
Callout

Loading a BAM file’s index

IGV automatically finds a BAM file’s index (.bai) as long as it is named correctly (e.g. NA12878.SLX.sample.bam.bai) and sits in the same folder as the BAM file. Do not load the .bai file directly — only load the .bam file.

Type snp1 into the search box and click Go.

IGV showing the loaded alignment track and search box with snp1
Loading the SNP BED file and BAM alignment track, then navigating to a named candidate SNP locus.

You should see individual aligned reads stacked as grey bars, with a red or blue bar above them summarizing the coverage and any mismatches at each position (the coverage track).

Sort reads by base to spot a real mismatch

First, click and drag to position the mismatched bases between the two vertical center guidelines. Then right-click (Mac: control-click) anywhere in the aligned reads and select Sort alignments by > base.

Mouse over (or click) the colored bar in the coverage track to see the exact allele counts and frequencies at that position.

Sorted alignments and allele count popup
Sorting reads by base and inspecting allele counts in the coverage track.

Observe the distribution of mismatches at this locus, and the lack of other mismatches nearby. This pattern — a consistent mismatch across many reads, at roughly the frequency you would expect for a heterozygous site, with no other noise around it — is a good sign of a real heterozygous SNP.

Challenge

Compare the two candidate SNPs

Go to the locus of the second candidate SNP by typing snp2 into the search box and clicking Go. Compare what you see to snp1 — in particular, look at the color/brightness of the mismatched bases (a fainter color indicates a lower base quality). Right-click the aligned reads and toggle Shade base by quality off and on to compare.

At snp2, many of the mismatched bases are drawn faintly, meaning IGV is shading them by low base call quality. Turning off Shade base by quality makes every mismatch look equally solid, which can be misleading — with shading on, it becomes clear that this locus is dominated by low-quality base calls rather than a strong, consistent signal like at snp1. This is a warning sign that snp2 may not be a real variant.

Sort and color reads by strand

Right-click anywhere in the aligned reads and select Sort alignments by > read strand, then right-click again and select Color alignments by > read strand.

Alignments colored and sorted by read strand
Reads sorted and colored by strand: an even split between the two colors is expected for a non-strand-specific library.

Observe where the mismatches fall relative to strand. If a sequencing library was not strand-specific, you would expect roughly a 50/50 split of reads on each strand, and a real variant should appear consistently on reads from both strands. If a mismatch only ever appears on reads from one strand, that is a red flag for a strand-specific artifact rather than a genuine variant.

Challenge

Decide: real SNP or artifact?

Using strand sorting/coloring, examine snp2 again. Given that this sequencing library was not strand-preserving (so you would expect the mismatch on both strands roughly equally), what do you conclude about snp2?

If the mismatched bases at snp2 are concentrated on reads from only one strand, despite the library not being strand-specific, this locus is likely a false positive rather than a real heterozygous SNP — combined with the low base quality observed earlier, there are now two independent pieces of evidence against it being real.

Using BLAT to check a suspicious signal

Sorting and coloring reads tells you a lot, but sometimes a coverage peak or a cluster of mismatches is suspicious for a different reason: the reads underneath it might not belong there at all. BLAT (Tools > BLAT…) lets you take a read’s sequence (or any feature, or a sequence you type in) and search a BLAT server for alternative places it could align in the genome.

A narrow coverage peak circled for investigation
An unusually narrow coverage spike — worth checking whether it is a unique alignment or a mapping artifact.

Right-click on a read at the locus you want to check and select BLAT read sequence.

Right-click context menu with Blat read sequence highlighted
Right-click a read and select “Blat read sequence” from the alignment track’s context menu.

BLAT returns every place in the genome the sequence aligns well, ranked by score.

BLAT results table showing many alternative alignments with similar scores
BLAT results for one read: many alternative alignments across different chromosomes with similar scores.

If BLAT returns one clearly best-scoring hit at the expected locus, the alignment is trustworthy. If it instead returns many alternative alignments with similar scores scattered across the genome — as in the example above — the read comes from a repetitive sequence that maps almost equally well to many locations, and the signal at your original locus should be treated with suspicion rather than taken as strong evidence for a variant.

Challenge

When would BLAT change your conclusion?

You already found two independent warning signs against snp2 being a real variant (low base quality, strand bias). Suppose BLAT on a snp2 read had instead returned dozens of equally good alternative alignments elsewhere in the genome. Would that change your confidence, and why?

It would add a third, independent line of evidence against snp2: reads that multi-map almost equally well to many locations cannot be confidently placed at any single locus, so any “variant” called from them is likely an artifact of ambiguous mapping rather than a true difference from the reference genome at that specific position.

Callout

Watch: IGV Sequencing Data Basics

Key Points
  • Aligned reads are stored in SAM/BAM format; recognizing terms like insert size, mapping quality, and CIGAR string helps you interpret what IGV shows you.
  • Load a BAM file with File > Load from File…; IGV finds its .bai index automatically as long as it is named correctly and kept alongside the BAM file.
  • Sorting and coloring aligned reads (by base, by base quality shading, or by read strand) helps you distinguish a real variant from a sequencing or alignment artifact.
  • Warning signs of a false-positive variant include low base quality and a mismatch that only appears on reads from one strand when it should not.
  • Tools > BLAT… (or right-click a read and choose BLAT read sequence) checks whether a read aligns uniquely or is a multi-mapping artifact from a repetitive region.