Module 2: Mapping & Variant Calling

Alignment forces every read to a position. Some of those positions are wrong.

Learning objectives

  1. Explain what a mapper does and where it fails.
  2. Distinguish coverage from depth.
  3. Read a VCF and interpret the key fields.
  4. Explain why HBV’s circular genome complicates mapping.

What mapping does

A mapper finds the most likely origin of each read in a reference genome. “Most likely” is the operative phrase — it always returns an answer, including for reads that came from somewhere else entirely.

reference  ...GCTGCTATGCCTCATCTTCTTGTTGG...
read              ATGCCTCATGTTCTT
                           ^ mismatch: variant, or error?

That question is what variant calling exists to answer.

Coverage vs depth

These get used interchangeably and should not be.

  • Depth at a position = how many reads overlap it.
  • Coverage = what fraction of the genome has any reads at all.
ImportantMean depth hides everything

“500× mean depth” is compatible with half the genome at zero. Amplicon data is especially prone to this: primer regions get enormous depth, inter-amplicon regions get none. Always plot depth along the genome. The mean is a summary that conceals exactly the problem you need to see.

Reading a VCF

#CHROM      POS  REF ALT QUAL  INFO
NC_003977.2 1896 G   A   2340  DP=1523;AF=0.98
  • POS is 1-based, relative to that specific reference. A different reference gives different coordinates for the same biology.
  • QUAL is confidence the variant is real — not that it matters.
  • DP is depth. AF is allele frequency.

An AF of 0.98 says nearly all reads carry the variant. An AF of 0.15 says 15% do, which is either a minority variant or an artifact — and telling those apart is the hard part.

Circular genomes

HBV’s genome is circular, and reference files are linear. A read spanning the artificial breakpoint will fail to map, producing an apparent coverage gap that is purely an artifact of file format.

WarningThe origin artifact

If you see a coverage dip at position 1, suspect the breakpoint before you suspect biology. Handle it by rotating the reference or using a circularity-aware mapper.

Exercises

  1. Map a workshop FASTQ. What fraction of reads mapped? Where did the rest go?
  2. Plot depth across the genome. Is it uniform? What does that tell you about library prep?
  3. Take a variant at AF=0.15. Real, or artifact? What evidence would decide it?
  4. Compare variant calls at depth thresholds 10 and 100. What changed?

Takeaways

  • A mapper always returns an answer. That is not the same as a correct one.
  • Plot depth; never trust the mean.
  • Coordinates are meaningless without the reference.