Interpreting Sequencing Data — Interpret sequencing/variant call outputs conceptually (intro) - Genetics

SECTION 1

Historical Context & Motivation

For most of human history, DNA was invisible and mysterious. Scientists knew that traits were inherited from parents to children, but they had no way to read the actual code hidden inside cells. The breakthrough came when researchers figured out how to sequence DNA — that is, to determine the exact order of the chemical letters (A, T, C, and G) that make up our genetic instructions. Once scientists could read DNA, they needed tools and formats to organize, display, and interpret the massive amounts of data that sequencing produces.

1977 • Sanger Sequencing Invented

Frederick Sanger developed the first practical method for reading DNA sequences one base at a time. This method could read a few hundred bases in a single run — slow by today's standards, but revolutionary at the time.

2003 • Human Genome Project Completed

After 13 years of international collaboration and roughly $2.7 billion, scientists published the first complete map of the human genome — all 3 billion base pairs. This project created the reference genome that researchers still compare new sequences against today.

2005 • Next-Generation Sequencing Arrives

Next-generation sequencing (NGS) technologies allowed millions of DNA fragments to be read at the same time. This made sequencing thousands of times faster and much cheaper.

2010s • Variant Calling Becomes Standard

Software tools were developed to automatically compare a patient's DNA sequence to the reference genome and flag differences called variants. These variant call outputs became a key part of genetics research and medicine.

2020s • Affordable Personal Genomics

Whole-genome sequencing now costs under $200, making it possible for doctors to use sequencing data to diagnose diseases and guide treatment for individual patients.

Today, sequencing machines produce enormous files filled with letters and numbers. The central question this lesson addresses is: How do we read and make sense of sequencing and variant call outputs? Understanding these data formats is the first step toward using genetics to solve real-world problems in medicine, agriculture, and forensics.

SECTION 2

Core Principles & Definitions

Before diving into data outputs, you need to understand a few foundational ideas. Sequencing data is built on a simple concept: your DNA is a long string of four chemical bases — adenine (A), thymine (T), cytosine (C), and guanine (G). A sequencing machine reads these bases and records them in order. But no machine reads the entire genome in one pass. Instead, the DNA is broken into short overlapping fragments called reads. Software then lines these reads up against a reference genome, like fitting puzzle pieces onto a picture on the box.

Read

A short segment of DNA sequence produced by a sequencing machine, typically 75–300 bases long. Think of it as one puzzle piece from the full genome picture.

Reference Genome

A standard, published DNA sequence that serves as the "correct answer key." New sequences are compared against it to find differences.

Alignment

The process of matching each read to its correct position on the reference genome. Software figures out where each puzzle piece belongs.

Variant

Any position where the sample's DNA differs from the reference genome. Variants include single-letter changes (SNPs), insertions, and deletions.

Quality Score

A number that tells you how confident the machine is that it read each base correctly. Higher scores mean greater confidence — like a test grade for accuracy.

✦

KEY TAKEAWAY

Think of DNA sequencing like using a paper shredder in reverse. The genome is a very long book. The sequencer shreds the book into tiny strips (reads), and then a computer program re-assembles the strips by matching overlapping words to rebuild the original text. A variant call is like noticing a typo — a spot where your copy of the book differs from the published edition.

SECTION 3

Visual Explanation — From Reads to Variants

The diagram below shows how raw sequencing reads are aligned to a reference genome. Notice that several reads overlap the same region. Where all the reads agree with the reference, the bases are shown in their normal color. Where one or more reads show a different letter, a variant is flagged.

Five sequencing reads are aligned to a reference genome. At position 10, three reads show a T instead of the expected C. This consistent difference across multiple reads is flagged as a variant — specifically a single nucleotide polymorphism (SNP).

In the diagram above, you can see the power of having multiple reads covering the same region. If only one read showed the T, it might just be a machine error. But because three out of five reads agree on the change, the software calls it a real variant with much higher confidence. This concept — requiring multiple reads to confirm a finding — is called read depth or coverage. More coverage means more confidence.

SECTION 4

How Variant Calling Works — The Pipeline

Sequencing data doesn't arrive ready to interpret. It goes through a series of processing steps, often called a bioinformatics pipeline. Think of it like an assembly line in a factory — raw materials go in one end, and a finished product comes out the other. Each step cleans, organizes, or analyzes the data a little further.

The Four Main Steps

Step 1 — Sequencing & Raw Data: The machine reads millions of short DNA fragments and outputs them in a file format called FASTQ. Each entry in a FASTQ file contains the sequence of bases and a quality score for each base. The quality score is a number that tells you how confident the machine is about each letter it read.

Step 2 — Alignment: Software takes each read and finds where it best fits on the reference genome, like matching puzzle pieces. The output is saved in a BAM file (Binary Alignment Map), which stores each read's position and how well it matched.

Step 3 — Variant Calling: Another program scans through all the aligned reads, compares them to the reference, and identifies positions where the sample differs. The output is stored in a VCF file (Variant Call Format) — a structured table listing every detected variant.

Step 4 — Annotation & Interpretation: Each variant is annotated with information like which gene it falls in, whether it changes a protein, and whether it has been seen before in databases. This is the step where data becomes meaningful biological information.

VARIANT ALLELE FREQUENCY

VAF = (Reads supporting variant ÷ Total reads at that position) × 100%

VAF = Variant Allele Frequency, the percentage of reads that show the alternate base. A VAF near 50% suggests a heterozygous variant (one copy changed). A VAF near 100% suggests a homozygous variant (both copies changed).

QUALITY SCORE (PHRED)

Q = −10 × log₁₀(P)

Q = quality score (higher is better). P = probability of an error. For example, Q = 30 means a 1 in 1,000 chance of error (99.9% accuracy). Q = 20 means a 1 in 100 chance (99% accuracy).

SECTION 5

Key File Formats — FASTQ, BAM, and VCF

Understanding sequencing data means knowing the three main file formats you'll encounter. Each format serves a different purpose in the pipeline. The diagram below shows a simplified view of what each file looks like and what information it contains.

The three main file formats in a sequencing pipeline: FASTQ holds raw reads and quality, BAM stores aligned positions, and VCF records the final list of variants.

File Format	Full Name	Contains	Typical Size
FASTQ	FASTQ (quality-scored FASTA)	Raw base sequences + quality scores for each base	50–200 GB per whole genome
BAM	Binary Alignment Map	Reads mapped to reference positions with alignment details	30–100 GB per whole genome
VCF	Variant Call Format	List of positions where the sample differs from reference	50–500 MB (much smaller)

Comparison of the three primary sequencing file formats

SECTION 6

Worked Example — Reading a VCF Entry

Let's walk through how to read a real variant call from a VCF file. Imagine you are a genetics researcher and you receive the following VCF entry for a patient. Your job is to figure out what it means.

ℹ

SAMPLE VCF LINE

chr7 117199646 rs113488022 G A 99 PASS DP=85;AF=0.47 GT 0/1

Interpreting the VCF Entry Step by Step

1
Step 1 — Identify the Location

The first two fields tell us the chromosome and position. This variant is on chromosome 7 at position 117,199,646. Think of this like a street address for the variant — it tells us exactly where on the genome to look.

Location: chr7, position 117,199,646
2
Step 2 — Check the Variant ID

The third field, rs113488022, is a known variant ID from the dbSNP database. The "rs" prefix means it has been seen and catalogued before. This is a well-studied variant in the CFTR gene, which is linked to cystic fibrosis.

Known variant: rs113488022 (in CFTR gene)
3
Step 3 — Read the Base Change

The fourth and fifth fields show the reference allele (G) and the alternate allele (A). This means the reference genome has a G at this position, but this patient's DNA shows an A instead. This is a single nucleotide polymorphism (SNP).

Base change: G → A (SNP)
4
Step 4 — Evaluate Quality and Depth

The quality score is 99, which is very high — it means the software is extremely confident this variant is real and not a sequencing error. The filter field says PASS, meaning it cleared all quality checks. The info field shows DP=85, meaning 85 reads covered this position, and AF=0.47, meaning about 47% of reads showed the alternate base A.

High confidence: Q=99, PASS, 85× coverage, 47% VAF
5
Step 5 — Determine the Genotype

The last field shows the genotype: 0/1. In VCF notation, 0 represents the reference allele and 1 represents the alternate allele. So 0/1 means this patient has one copy of the reference (G) and one copy of the alternate (A). This is a heterozygous variant — consistent with the ~47% VAF we calculated above.

Genotype: 0/1 → heterozygous (one G, one A)

✦

KEY TAKEAWAY

Reading a VCF line is like reading a medical chart. Each column gives you a different piece of the story: where the difference is (address), what changed (base swap), how confident we are (quality score), and whether one or both copies of the gene are affected (genotype). With practice, you can scan a VCF file just like a doctor reads a chart.

SECTION 7

Types of Variants & Their Significance

Not all variants are created equal. Some are harmless differences that make each person unique, while others can cause serious diseases. Understanding the different types of variants and their potential effects is essential for interpreting sequencing data in a meaningful way.

Variant Type	What It Looks Like	Example	Possible Effect
SNP (Single Nucleotide Polymorphism)	One base is swapped for another	Reference: A → Sample: G	May change a protein, be harmless, or cause disease
Insertion	Extra bases are added	Reference: ATG → Sample: ATCG	Can shift the reading frame, often disrupting a protein
Deletion	Bases are removed	Reference: ATGC → Sample: ATC	Can shift the reading frame or remove part of a gene
Indel	Insertion and deletion combined	Complex rearrangement at one location	Effects depend on size and location
Structural Variant	Large-scale rearrangement (>50 bases)	Large deletion, duplication, or inversion	Can remove or duplicate entire genes

Common variant types found in sequencing data

Impact Categories

Variant Impact Spectrum

Harmless Disease-causing

✦

KEY TAKEAWAY

Think of variants like typos in a recipe. Some typos don't matter at all ("add sugar" vs. "add suger" — you still understand). Some are confusing but manageable ("add 2 cups" vs. "add 3 cups"). But others can ruin the dish entirely ("add salt" vs. "add soap"). Geneticists classify variants from benign to pathogenic to figure out which "typos" actually matter for a patient's health.

SECTION 8

Connection to Advanced Genomic Analysis

The basic interpretation skills you've learned in this lesson are the foundation for much more advanced analyses. As you progress in genetics, you'll encounter larger datasets, more complex variant types, and sophisticated tools for understanding what variants mean for human health and evolution.

Introductory Concept (This Lesson)	Advanced Application
Reading a single VCF entry	Filtering thousands of variants to find disease-causing mutations using population databases like gnomAD
Understanding SNPs	Genome-wide association studies (GWAS) that link SNPs to complex traits like height or diabetes risk
Quality scores for individual bases	Machine learning algorithms that evaluate variant quality across millions of positions simultaneously
Heterozygous vs. homozygous genotypes	Inheritance pattern analysis for genetic counseling (autosomal dominant, recessive, X-linked)
Variant allele frequency in one sample	Tumor mutation analysis where mixed populations of normal and cancer cells create complex allele frequencies

How today's concepts connect to advanced genomic analysis

One exciting frontier is pharmacogenomics — using a patient's sequencing data to choose the best medication and dose. Some variants in drug-processing genes mean that a standard dose could be too strong or too weak for certain patients. By reading VCF data, doctors can personalize treatment plans. Another growing area is prenatal screening, where sequencing tiny amounts of fetal DNA floating in a pregnant person's blood can reveal chromosomal conditions early in pregnancy.

💡

LOOKING AHEAD

As sequencing costs continue to drop and artificial intelligence improves variant interpretation, reading sequencing data will become as routine in medicine as reading blood test results. The concepts you're learning now will be essential skills in biology, medicine, and biotech careers.

SECTION 9

Practice Problems

Question 1: CONCEPTUAL

A sequencing machine produces a FASTQ file. In your own words, what two key pieces of information does a FASTQ file contain for each read, and why are both important?

Show answer

A FASTQ file contains the DNA sequence (the string of A, T, C, and G bases that were read) and a quality score for each base. The sequence tells you what was read, while the quality score tells you how confident the machine is about each letter. Both pieces are important because a read with many low-quality bases might contain errors that could be mistaken for real variants. Without quality scores, we wouldn't be able to tell the difference between a true genetic variant and a machine mistake.

Question 2: BASIC CALCULATION

At a certain position on chromosome 3, a total of 60 reads are aligned. Of those, 42 reads show a T (which matches the reference), and 18 reads show a C. Calculate the variant allele frequency (VAF) for the C allele. Based on the VAF, is this variant likely homozygous or heterozygous?

Show answer

VAF = (Reads supporting variant ÷ Total reads) × 100% = (18 ÷ 60) × 100% = 30%. A VAF of 30% is closer to 50% than to 0% or 100%, which suggests this could be a heterozygous variant (0/1 genotype). However, 30% is somewhat below the expected 50% for a perfect heterozygous call. This might occur due to sampling variation, mapping biases, or slightly unequal representation of the two alleles. The variant caller software would consider this along with other quality metrics before making a final genotype call.

Question 3: INTERMEDIATE

You receive the following VCF line: chr17 7577121 . G GA 45 PASS DP=30;AF=0.50 GT 0/1 Identify: (a) the type of variant, (b) the genotype, (c) the read depth, and (d) one concern you might have about this variant call.

Show answer

(a) This is an insertion because the alternate allele (GA) is longer than the reference allele (G) — an extra A base has been added. (b) The genotype is 0/1, meaning heterozygous — the patient has one normal copy (G) and one copy with the insertion (GA). (c) The read depth (DP) is 30, meaning 30 reads covered this position. (d) A concern is that DP=30 is relatively low coverage. While it passed the quality filter, higher coverage (such as 50× or more) would give greater confidence. Also, the quality score of 45 is decent but not as high as the Q=99 we saw in our worked example, so there is slightly more uncertainty about this call.

Question 4: APPLIED

A doctor is examining a patient with symptoms of a genetic disease. Sequencing reveals 4.2 million variants compared to the reference genome. The doctor needs to narrow this list down to find the disease-causing variant. Describe at least three filtering strategies the doctor could use to reduce 4.2 million variants to a manageable shortlist.

Show answer

The doctor could use several strategies: (1) Filter by quality — remove any variants with low quality scores or low read depth, since these might be sequencing errors rather than real variants. (2) Filter by frequency — remove common variants seen in large population databases (like gnomAD). If a variant appears in 30% of all people, it is unlikely to cause a rare disease. Disease-causing variants are usually rare (less than 1% frequency). (3) Filter by location — focus on variants that fall within genes, especially in exons (the protein-coding regions), since variants in non-coding regions are less likely to affect protein function. (4) Filter by predicted impact — use annotation tools that predict whether a variant changes a protein's amino acid sequence. Synonymous (silent) variants that don't change the protein are usually less important. (5) Filter by gene — if the disease is known to be linked to specific genes, the doctor can focus only on variants in those genes.

Question 5: CRITICAL THINKING

Two patients are sequenced at the same genomic position. Patient A has a VCF genotype of 0/1 with VAF = 48% and DP = 120. Patient B has a VCF genotype of 0/1 with VAF = 15% and DP = 20. Both entries say PASS in the filter column. Which variant call would you trust more, and why? Could there be a biological (not technical) explanation for Patient B's unusual VAF? Explain your reasoning.

Show answer

Patient A's variant call is far more trustworthy. With 120 reads of coverage and a VAF of 48% — very close to the expected 50% for a heterozygous variant — the data strongly supports a true heterozygous SNP. Patient B's call is concerning for two reasons: the low read depth of 20 means random sampling variation has a larger effect, and the VAF of 15% is much lower than the expected 50% for a heterozygous variant. There are two possible explanations. The technical explanation is that 20× coverage simply isn't enough to get an accurate VAF measurement, and random chance made the ratio look skewed. The biological explanation is more interesting: if Patient B's sample includes a mixture of cell types — for example, a tumor biopsy containing both normal cells and cancer cells — then a VAF of 15% could mean the variant exists only in the cancer cells, which make up roughly 30% of the sample (since a heterozygous variant in 30% of cells would show ~15% VAF). This kind of analysis is central to cancer genomics, where "somatic mosaicism" creates mixed populations with unusual allele frequencies.

SUMMARY

Lesson Summary

Summary

DNA sequencing machines break genomes into millions of short reads and record them with quality scores in FASTQ files. Software then aligns these reads to a reference genome, producing BAM files. A variant caller compares the aligned reads to the reference and outputs a VCF file listing every position where the sample's DNA differs — these differences are called variants. Each VCF entry records the chromosome, position, reference base, alternate base, quality score, and genotype (0/0 for homozygous reference, 0/1 for heterozygous, 1/1 for homozygous alternate).

Key metrics for evaluating variant calls include read depth (DP) — the number of reads covering a position — and variant allele frequency (VAF), the fraction of reads showing the alternate base. Variants range from benign (harmless) to pathogenic (disease-causing), and include types like SNPs, insertions, deletions, and structural variants. These foundational skills in reading sequencing data connect directly to advanced applications in personalized medicine, cancer genomics, and genetic counseling.