The Gold Standard of Genetics | Long-Read Sequencing, Repeat Expansions, Structural Variants, Phasing and Orthogonal Confirmation

THE GOLD STANDARD OF GENETICS · SPECIALIST STUDY

Choose long reads for the biological question, then prove the answer survives validation

Long reads can span repeats, connect distant variants, reveal structural rearrangements and support haplotype-resolved analysis that short fragments may leave ambiguous. They do not remove sample bias, platform error, mapping uncertainty, reference limitations or interpretation burden. A trustworthy workflow begins with the diagnostic or research question, makes read length and accuracy visible, separates primary observations from assembly and caller inference, benchmarks the hard variant classes, and confirms consequential findings with an independent method or evidence stream suited to the mechanism.

This guide belongs to the Genetics subject root, the Gold Standard of Education, and the How X Works hub. It begins with a concrete decision, then tests whether the method still holds when the context changes.

The whole problem at a glance

Each transformation can add resolution and error; the report should make both visible.

CHAPTER 2 OF 13 · Match the platform to the question

2. Separate raw signal, base calls, reads and interpretation

Preserve the transformations from instrument signal to base-called reads, alignment or assembly, variant calls and clinical interpretation. Each layer has its own version and error model. The working unit here is one specimen, one sequencing run and pipeline version, one genomic question, and one reported claim with a stated validation boundary. State whose sample, memory, decision or record is being examined, what time boundary applies and which claim is still outside the evidence.

Worked example: A basecaller update changes the apparent homopolymer sequence. The laboratory can reproduce the earlier result because raw data, model and pipeline versions were retained. Put the starting information, decision, later observation and consequence on separate lines. That chronology stops a coherent story from replacing a reproducible account.

Evidence to retain: raw signal; basecaller and model; read files; quality scores; alignment; reference; assembly; caller; filters; interpretation version Attach dates, versions, provenance, custody and the decision use. Record unavailable evidence explicitly; absence is not a negative result and cannot safely be turned into reassurance.

Boundary and misconception: A polished consensus can hide uncertainty that remains obvious in individual reads. Test a credible competing explanation and name what evidence would change the conclusion. The pass condition is a bounded, reviewable claim, not a promise that the method removes uncertainty.

Evidence discipline: record the observation, version or time boundary that supports this chapter’s conclusion. Preserve the rejected alternative as well as the selected action. That makes a later review able to distinguish a changed world from an undocumented change of mind.

Changed-context transfer: Hold the specimen fixed and update only the basecaller; identify which comparisons are needed before accepting changed calls. The method transfers only if the decision question and measurement meaning survive the change. If either changes, rebuild the comparison rather than copying the earlier threshold or weight.

Contents · Previous chapter · Next chapter

CHAPTER 4 OF 13 · Resolve repeats, structures and haplotypes

4. Resolve structural variants with more than one signature

Combine split reads, discordant alignment, depth, assembly and breakpoint context rather than relying on one caller label. Represent complex events without forcing them into a misleading simple deletion. The working unit here is one specimen, one sequencing run and pipeline version, one genomic question, and one reported claim with a stated validation boundary. State whose sample, memory, decision or record is being examined, what time boundary applies and which claim is still outside the evidence.

Worked example: Reads support an inversion with a nested deletion. The team reviews breakpoints and local assembly before choosing a representation that downstream interpretation can understand. Put the starting information, decision, later observation and consequence on separate lines. That chronology stops a coherent story from replacing a reproducible account.

Evidence to retain: supporting reads; strand and alignment; depth; local assembly; breakpoint sequence; alternate contigs; caller versions; manual review Attach dates, versions, provenance, custody and the decision use. Record unavailable evidence explicitly; absence is not a negative result and cannot safely be turned into reassurance.

Boundary and misconception: Different callers may describe the same event differently, and apparent disagreement can be representational rather than biological. Test a credible competing explanation and name what evidence would change the conclusion. The pass condition is a bounded, reviewable claim, not a promise that the method removes uncertainty.

Evidence discipline: record the observation, version or time boundary that supports this chapter’s conclusion. Preserve the rejected alternative as well as the selected action. That makes a later review able to distinguish a changed world from an undocumented change of mind.

Changed-context transfer: Move the breakpoint into a segmental duplication and identify which mapping claims weaken. The method transfers only if the decision question and measurement meaning survive the change. If either changes, rebuild the comparison rather than copying the earlier threshold or weight.

Contents · Previous chapter · Next chapter

CHAPTER 6 OF 13 · Work and benchmark the pipeline

6. Use phasing to answer a defined cis–trans question

Distinguish read-backed phase, pedigree phase, statistical phase and assembly haplotypes. State distance, informative markers and confidence. The working unit here is one specimen, one sequencing run and pipeline version, one genomic question, and one reported claim with a stated validation boundary. State whose sample, memory, decision or record is being examined, what time boundary applies and which claim is still outside the evidence.

Worked example: Two pathogenic variants appear on the same long reads, supporting cis phase. Parental testing confirms the haplotype before the result is used for recessive-disease counselling. Put the starting information, decision, later observation and consequence on separate lines. That chronology stops a coherent story from replacing a reproducible account.

Evidence to retain: read links; mapping quality; informative heterozygotes; pedigree genotypes; recombination; haplotag method; phase block length Attach dates, versions, provenance, custody and the decision use. Record unavailable evidence explicitly; absence is not a negative result and cannot safely be turned into reassurance.

Boundary and misconception: Long reads improve physical linkage but cannot phase every locus when reads do not span informative markers or mapping is ambiguous. Test a credible competing explanation and name what evidence would change the conclusion. The pass condition is a bounded, reviewable claim, not a promise that the method removes uncertainty.

Evidence discipline: record the observation, version or time boundary that supports this chapter’s conclusion. Preserve the rejected alternative as well as the selected action. That makes a later review able to distinguish a changed world from an undocumented change of mind.

Changed-context transfer: Double the distance between variants and determine whether read-backed phase still holds. The method transfers only if the decision question and measurement meaning survive the change. If either changes, rebuild the comparison rather than copying the earlier threshold or weight.

Contents · Previous chapter · Next chapter

CHAPTER 8 OF 13 · Work and benchmark the pipeline

8. Work a complete difficult-locus case

Join prior negative testing, sample QC, long-read evidence, repeat or structural analysis, phase, phenotype and confirmation in a single chronology. The working unit here is one specimen, one sequencing run and pipeline version, one genomic question, and one reported claim with a stated validation boundary. State whose sample, memory, decision or record is being examined, what time boundary applies and which claim is still outside the evidence.

Worked example: A fictional patient has a suspected expansion disorder. High-molecular-weight DNA yields spanning reads with an interrupted expansion; a targeted assay confirms the size range, and segregation supports relevance. Put the starting information, decision, later observation and consequence on separate lines. That chronology stops a coherent story from replacing a reproducible account.

Evidence to retain: prior testing; phenotype; sample and library QC; read support; motif; phase; caller and reference; confirmation; family evidence; report limits Attach dates, versions, provenance, custody and the decision use. Record unavailable evidence explicitly; absence is not a negative result and cannot safely be turned into reassurance.

Boundary and misconception: The example shows the chain, not a diagnosis. Read count, repeat size and phenotype must not be collapsed into deterministic prognosis. Test a credible competing explanation and name what evidence would change the conclusion. The pass condition is a bounded, reviewable claim, not a promise that the method removes uncertainty.

Evidence discipline: record the observation, version or time boundary that supports this chapter’s conclusion. Preserve the rejected alternative as well as the selected action. That makes a later review able to distinguish a changed world from an undocumented change of mind.

Changed-context transfer: Change the expansion to a complex structural rearrangement and replace each evidence step that no longer fits. The method transfers only if the decision question and measurement meaning survive the change. If either changes, rebuild the comparison rather than copying the earlier threshold or weight.

Contents · Previous chapter · Next chapter

Release and reporting gate

A negative result also needs a stated callable region and limitation.

CHAPTER 10 OF 13 · Confirm, interpret and report

10. Choose orthogonal confirmation that tests the mechanism

Match confirmation to the claim: targeted sequencing, PCR-free repeat assay, optical mapping, karyotype, microarray, segregation, RNA or another independent evidence stream. The working unit here is one specimen, one sequencing run and pipeline version, one genomic question, and one reported claim with a stated validation boundary. State whose sample, memory, decision or record is being examined, what time boundary applies and which claim is still outside the evidence.

Worked example: An insertion call is confirmed with a breakpoint-spanning assay and family segregation rather than by rerunning the same caller on the same reads. Put the starting information, decision, later observation and consequence on separate lines. That chronology stops a coherent story from replacing a reproducible account.

Evidence to retain: claimed variant; independent chemistry or measurement; breakpoint or motif; positive and negative controls; sample identity; segregation; confirmation limit Attach dates, versions, provenance, custody and the decision use. Record unavailable evidence explicitly; absence is not a negative result and cannot safely be turned into reassurance.

Boundary and misconception: Orthogonal does not merely mean different software applied to the same signal, and no single method confirms every structural claim. Test a credible competing explanation and name what evidence would change the conclusion. The pass condition is a bounded, reviewable claim, not a promise that the method removes uncertainty.

Evidence discipline: record the observation, version or time boundary that supports this chapter’s conclusion. Preserve the rejected alternative as well as the selected action. That makes a later review able to distinguish a changed world from an undocumented change of mind.

Changed-context transfer: Replace an insertion with a methylation claim and identify what independent evidence would be meaningful. The method transfers only if the decision question and measurement meaning survive the change. If either changes, rebuild the comparison rather than copying the earlier threshold or weight.

Contents · Previous chapter · Next chapter

CHAPTER 12 OF 13 · Govern change and transfer

12. Report negative findings with callable scope

State which loci, repeat sizes, structural classes, copy states and genomic regions were adequately assessed. Include sample, coverage and pipeline limitations. The working unit here is one specimen, one sequencing run and pipeline version, one genomic question, and one reported claim with a stated validation boundary. State whose sample, memory, decision or record is being examined, what time boundary applies and which claim is still outside the evidence.

Worked example: A report says no reportable expansion was detected within the validated size and motif range; it does not say the genome contains no relevant variant. Put the starting information, decision, later observation and consequence on separate lines. That chronology stops a coherent story from replacing a reproducible account.

Evidence to retain: callable regions; coverage distribution; validated classes and ranges; excluded loci; quality failures; prior probability; alternative testing options Attach dates, versions, provenance, custody and the decision use. Record unavailable evidence explicitly; absence is not a negative result and cannot safely be turned into reassurance.

Boundary and misconception: A negative whole-genome label can sound comprehensive while hiding uncallable repeats, paralogs or complex rearrangements. Test a credible competing explanation and name what evidence would change the conclusion. The pass condition is a bounded, reviewable claim, not a promise that the method removes uncertainty.

Evidence discipline: record the observation, version or time boundary that supports this chapter’s conclusion. Preserve the rejected alternative as well as the selected action. That makes a later review able to distinguish a changed world from an undocumented change of mind.

Changed-context transfer: Lower locus coverage below validation and rewrite the negative statement without false reassurance. The method transfers only if the decision question and measurement meaning survive the change. If either changes, rebuild the comparison rather than copying the earlier threshold or weight.

Contents · Previous chapter · Next chapter

Limits and misconceptions

Long reads do not remove error

They change the evidence available and introduce platform-, model- and context-specific error.

Genome-wide accuracy can hide local failure

Benchmark the variant class and difficult region that matter to the question.

One confirmation method does not fit every claim

Sequence, structure, phase, repeat and methylation claims require different evidence.

A negative result has a callable boundary

Report what the specimen, run and validated pipeline could actually assess.

Practise, verify and continue

Build a release checklist for one difficult locus: question, sample integrity, platform, read support, alignment or assembly, variant representation, phase, class-specific benchmark, confirmation, interpretation and negative boundary.

Apply the method to a pangenome-based analysis. Preserve raw-read provenance and class-specific validation, then rebuild coordinate representation, benchmarking, clinical databases and report portability for the new reference system.

Practical evidence check

Before implementation, write a one-page evidence map with the decision, unit of analysis, version or time boundary, inputs, outcomes, uncertainty, authority and stop rule. Attach one worked case that passes and one that fails. A reviewer should be able to reproduce the reasoning without relying on the memory of the person who built the system.

Run one changed-context rehearsal before approval. Alter a mechanism rather than a cosmetic detail, then identify which calculation, safeguard or governance gate must change. Record any conclusion that does not transfer. This short rehearsal is often the fastest way to expose a hidden assumption before it becomes an operational dependency.

Related eduKate reading

Phasing errors, identity-by-descent and kinship inference
Mosaicism, allele fractions and tissue-specific detection
Reverse phenotyping and genotype-first recall

Authoritative sources and scope

NHGRI workshop report on long-read human genome sequencing and applications
NIST Genome in a Bottle benchmarks and reference materials
NIST Platinum Pedigree long-read benchmark
NIST benchmark for germline large insertions and deletions

These sources support the visible distinctions and evidence practices described here. They do not replace patient-specific genetic counselling, laboratory validation, consent, local regulation or a clinician’s interpretation. All worked cases are fictional, recurrence figures are illustrative rather than personal estimates, and no reader should start, stop or change medical care from this educational article.

Back to contents · Return to the Genetics guide