Bioinformatics for the Clinical Lab

From raw reads to a report: FASTQ, FASTA, BAM, and VCF in order, Phred quality (Q30 = 99.9%), trimming, alignment, variant calling, annotation, and filtering; depth and variant allele fraction with a worked pileup; and the lab-operations side - genome builds, pipeline validation, and version control.

5 min
6 steps
4 questions
Lesson 32 of 60

In this lesson

The files, in order
Phred quality
The pipeline
Depth and VAF
Running a pipeline in a lab
What to take from this

Bioinformatics

The files, in order

File	Holds	Made by
FASTQ	raw reads, four lines each: name, bases, separator, a quality character per base	the sequencer (after demultiplexing)
FASTA	the reference genome sequence	a curated public reference
SAM / BAM	alignments: where each read maps and how well (BAM is the compressed binary form)	the aligner
VCF	variant calls: one per line with evidence and quality	the variant caller

¹

@RUN1:lane1:read00042
GATCACAGGTCTATCACCCTATTAACCAC
+
IIIIIIIIIFFFFFAAAA<<<<7777,,,

The fourth line encodes a quality score for each base: here high at the start and falling toward the end, as reads typically do.

Quick check

Which file holds the aligned reads?

Phred quality

Q = -10 × log₁₀(P), where P is the chance the base call is wrong ¹:

Q	Error rate	Accuracy
10	1 in 10	90%
20	1 in 100	99%
30	1 in 1,000	99.9%
40	1 in 10,000	99.99%

Run QC reports the percentage of bases at Q30 or above; most runs aim for 80% or more.

Quick check

A base has Phred quality Q30. What is the chance it’s wrong?

The pipeline

QC and trimming: remove adapter sequence and low-quality tails; discard poor reads ¹.
Alignment: place each read at its best match on the reference, tolerating real variants and errors; output BAM. Mark PCR duplicates.
Variant calling: at each position, decide statistically whether reads that disagree with the reference reflect a real variant - weighing depth, base and mapping quality, and strand balance ¹. Separate callers handle small variants, copy number, and structural variants.
Annotation: add gene, transcript, HGVS name, predicted effect, population frequency (gnomAD), and database entries (ClinVar, COSMIC) ².
Filtering and interpretation: drop low-quality, common, or irrelevant calls; a qualified analyst classifies the rest (ACMG/AMP for germline, AMP/ASCO/CAP tiers for somatic) and decides what is reported.

Depth and VAF

Forty sequencing reads stacked against a reference sequence; at one position eight reads carry T instead of the reference C, giving a variant allele fraction of 20 percent at 40x depth. A side panel lists what VAFs suggest - about 50 percent germline heterozygous, about 100 percent homozygous, 20 percent somatic or impure tumor, 1-5 percent subclonal. — A pileup at 40x depth. Eight variant reads give a VAF of 20%; interpreting it depends on whether the sample is germline or tumor. Credit: StudyCorner diagram after Buckingham, Molecular Diagnostics · CC BY 4.0 · Source

Depth (coverage) at a position: how many reads span it after duplicate removal. Reports state the minimum depth and the fraction of target bases covered (for example, 99% at 100x or more).
Variant allele fraction (VAF): variant reads ÷ total reads.

Interpretation depends on the sample:

VAF	Germline sample	Tumor sample
about 50%	heterozygous	possibly germline - confirm in normal tissue
about 100%	homozygous or hemizygous	germline, or a variant with loss of the other allele
10-40%	suspicious for mosaicism or artifact	typical somatic, scaled by tumor purity
1-5%	usually artifact	subclone; needs deep coverage and often UMIs

A tumor at 40% purity with a heterozygous clonal mutation would show a VAF near 20%. At low depth, VAF is meaningless: 1 of 4 reads is “25%” with huge uncertainty.

Quick check

A position has 500 reads; 25 carry a variant. What is the VAF?

Running a pipeline in a lab

Reference build: GRCh37 (hg19) and GRCh38 (hg38) give the same variant different coordinates. Record the build, and name variants against versioned transcripts.
Validation: a pipeline is validated like any assay - sensitivity and specificity for each variant type against reference materials - and revalidated when software, versions, or settings change.
Version control and locked configurations, so every report can be traced to exactly the code that produced it.
Storage and privacy: FASTQ and BAM files are huge and contain identifiable genomes; retention and security policies apply.
Known blind spots: low-complexity regions, pseudogenes, large indels, and GC-extreme regions; reports should state them.

Quick check

Why must a lab record which reference genome build its pipeline uses?

What to take from this

FASTQ (raw reads with qualities) → align to the FASTA reference → BAM → call → VCF → annotate and interpret. Phred Q30 means a 1-in-1,000 error. Depth counts reads at a base; VAF is variant reads over total, read in light of germline versus tumor and tumor purity. A clinical pipeline names its genome build, is validated and version-controlled, and states its blind spots.

Lesson complete

Nice work.

1day streak

0/1today's goal

–correct

Up next · 4 min

Melt-Curve Analysis and Epigenetic Detection

Next lesson

Back to course Today

Sources for this lesson

1
Lela Buckingham. Molecular Diagnostics: Fundamentals, Methods, and Clinical Applications. 3rd ed. F.A. Davis Company. 2019. verifiedThe standard clinical molecular-diagnostics textbook for MLS/MB programs; author holds MB DLM(ASCP). Covers nucleic-acid chemistry, techniques, lab operations, and applications across infectious disease, oncology, genetics, and identity. Primary topic reference for the ASCP MB program.
2
Bruce Alberts, Rebecca Heald, Alexander Johnson, David Morgan, Martin Raff, Keith Roberts, Peter Walter. Molecular Biology of the Cell. 7th ed. W. W. Norton & Company. 2022. verifiedThe canonical cell/molecular biology textbook; used for nucleic-acid chemistry and the central dogma.