The first test is a decision, not a default. The deeper principle is matching test scope to pre-test probability and to the variant classes the phenotype implies — every assay has a blind spot, and the art is choosing the one whose blind spots don't overlap your likeliest diagnosis. Five branch points operationalize that. The Diagnostic Yields module holds the full numbers; this section is the decision layer.

1. How specific is the phenotype? A tight, recognizable syndrome (e.g., classic Rett, tuberous sclerosis) points to single-gene testing or a focused panel — high pre-test probability means a narrow, deeply-covered assay is both faster and less likely to surface incidental uncertainty. A broad or undifferentiated presentation (unexplained GDD/ID) points to exome (or genome): when you cannot predict the gene, a panel's curated list becomes a liability, because the causal gene may simply not be on it. Exome interrogates the whole coding space and can be reanalyzed as gene-disease knowledge grows — a panel cannot.

2. Are both parents available? For severe, early-onset, de novo-enriched phenotypes (DEE, GDD with seizures, MCA), trio sequencing (proband + both parents) approximately doubles yield versus singleton — OR ~2.04 (Clark 2018) — and phases compound heterozygotes. The mechanism is twofold: trio data let the lab instantly flag a variant as de novo (a strong pathogenicity signal, ACMG PS2) and confirm that two recessive variants are in trans rather than on the same allele. When parents are available and the phenotype is severe and early-onset, default to trio.

3. Does the picture suggest a repeat expansion? Progressive ataxia, myotonic dystrophy, or a Huntington-like phenotype should trigger dedicated repeat testing (repeat-primed PCR, Southern blot, or long-read). Short-read WES/WGS does NOT reliably detect trinucleotide/pentanucleotide expansions, because reads shorter than the expanded tract cannot be uniquely aligned across it — the expansion is literally invisible to the platform, not merely missed.

4. MCA/dysmorphism or infantile spasms? CMA retains first-tier value: ~15–25% in multiple congenital anomalies, and 14% in infantile epileptic spasms (IESS) (vs. WES 26%). CMA's strength is resolution at the dosage level — it sees whole-gene and contiguous-gene deletions/duplications that a sequencing read, focused on single bases, can miss. See CNV Interpretation for what CMA detects (CNVs, aneuploidy, UPD/AOH on SNP arrays) and misses (SNVs, balanced rearrangements, repeat expansions).

5. How urgent is turnaround? An acutely ill NICU neonate needs rapid WGS/WES (35–50% yield; Maron JAMA 2023) — speed is the point, because diagnosis changes management in 38–50% of cases, and the window to redirect care is days, not weeks.

Bottom line: for most undifferentiated NDD/epilepsy with parents available, modern first-tier is trio exome (± concurrent CNV-seq); CMA stays first-tier for MCA/dysmorphism; targeted panels fit phenotypically tight presentations; repeat disorders and NICU urgency each demand their own pathway. The unifying question is always: what variant class is most likely here, and does my chosen test actually see it?

A Variant of Uncertain Significance (VUS) is exactly that: a variant for which the available evidence is insufficient to call it pathogenic or benign. The key insight is that uncertain is a statement about the evidence, not about the variant's true effect — the variant is already either disease-causing or not; we simply lack the data to know which. A VUS is therefore NOT a partial diagnosis, a 'weak positive,' or a hedge. How the lab arrives at this tier (the ACMG/AMP evidence codes) is covered in Variant Interpretation; this section is about what you do with one in the room.

The cardinal rule: clinical decisions are not made on a VUS. Under the ACMG/AMP 2015 standards (Richards et al. 2015), a VUS must not be used to confirm or exclude a diagnosis. Do not change management, initiate surveillance, alter treatment, or drive reproductive decisions on an unresolved VUS. Manage the patient on their clinical findings — as if the VUS had not been reported. The reasoning is asymmetric-harm: acting on a VUS that proves benign means surveillance, anxiety, or even reproductive decisions imposed for nothing, whereas waiting costs little because true pathogenic variants tend to declare themselves clinically and through reclassification.

Reclassification skews benign. When VUS are later reclassified, the majority move toward benign rather than pathogenic — an expected consequence of how variants enter the VUS bin: most are simply rare, and rarity alone is weak evidence, so as population databases grow many prove too common to cause a severe Mendelian disorder. A VUS is therefore not 'probably the answer,' and presenting it as the likely cause manufactures false certainty that can harm the family.

What to say vs. what not to say:

What can move a VUS over time: periodic lab reclassification as databases grow; segregation and parental (trio) testing — confirming de novo status is often the single most informative step, since a confirmed de novo variant in a constrained gene adds strong pathogenic weight (ACMG PS2), while finding the same variant in an unaffected parent pushes toward benign; deeper phenotyping; and functional or RNA studies that supply the experimental evidence the variant lacked. Set honest expectations about re-contact: responsibility for notifying you of a reclassification — lab versus ordering clinician — is not standardized. Do not promise the family an automatic system will reach them; the practical safeguard is a clinician-owned plan, not an assumed lab callback. Document the VUS with explicit uncertainty language and own the plan to revisit it (commonly reanalysis in 1–2 years).

A negative report is time-stamped, not permanent. It means no diagnostic variant was found with today's knowledge, on this platform, in this tissue — not that a genetic cause is excluded. The trap is reading absence of evidence as evidence of absence: a 'normal' result is bounded by three things the report rarely spells out — the maturity of gene-disease knowledge at the time, the variant classes the assay can physically detect, and the tissue that was sampled. Before declaring the workup done, run the phenotype past four under-the-radar blind spots, each with its own fix — see Diagnostic Yields for the reanalysis and platform-yield numbers.

The repeat-expansion blind spot — know the list. Standard short-read sequencing misses: Friedreich ataxia (FXN GAA), spinocerebellar ataxias (CAG), CANVAS (RFC1 AAGGG), FXTAS (FMR1 CGG premutation), myotonic dystrophy DM1/DM2, Huntington disease (HTT CAG), and C9orf72 (ALS/FTD). A progressive ataxia, a myotonic or Huntington-like picture, or unexplained adult-onset tremor/ataxia should send you to dedicated repeat testing regardless of a 'negative' exome.

Methylation and imprinting. Angelman and Prader-Willi syndromes — and other imprinting disorders — are driven by methylation defects or UPD that sequence reads alone cannot see, because the DNA letters are normal; what is abnormal is the parent-of-origin epigenetic mark layered on top. The SNRPN methylation test detects ~99% of PWS and ~80% of AS; the gap for Angelman reflects its mechanistic heterogeneity — the ~10–15% caused by UBE3A point variants have normal methylation, so a normal methylation result in a classic Angelman picture still prompts UBE3A sequencing. The Methylation & Imprinting module covers the parent-of-origin mechanics in detail.

Mosaicism and tissue choice. Somatic mosaic overgrowth (e.g., PIK3CA, AKT3) and brain-restricted variants causing focal cortical dysplasia or hemimegalencephaly may be absent from blood entirely. The route is deep-coverage sequencing of the affected tissue — skin, biopsy, or resected cortex — not a repeat of the blood test, as the Somatic Mosaicism module details.

Below the detection floor. Deep-intronic, regulatory, and structural variants below standard WES detection are next-step territory for genome sequencing or RNA-seq.

The clinical reflex: after every negative result, ask which of these blind spots fits this phenotype? before calling the workup complete.

A molecular diagnosis is not the end of the workup — it reframes management. The diagnosis converts an open-ended 'what is wrong?' into a defined problem with a known natural history, a known complication set, and sometimes a specific molecular target — which is precisely why pinning down the gene matters even when no cure exists.

Disease-modifying vs. symptomatic. A disease-modifying therapy alters the underlying biology or natural history; symptomatic/supportive care treats the manifestations without changing the course. The distinction is not academic: disease-modifying agents are usually most powerful early, before irreversible injury accrues, so the timing of diagnosis can determine whether the most effective window is still open. For a growing list of monogenic conditions, a precise molecular diagnosis is the gateway to disease-modifying treatment — see Therapies and Neuromuscular Disease. In SMA, SMN-restoring therapy (nusinersen, an intrathecal ASO; risdiplam, an oral SMN2 splicing modifier; onasemnogene abeparvovec, AAV9-SMN1 gene replacement) transforms the natural history and is most effective given presymptomatically. In DMD, exon-skipping ASOs (e.g., eteplirsen for exon-51-amenable deletions) restore the reading frame in eligible genotypes. None of these can be offered without the exact molecular result.

Precision symptomatic / metabolic treatment. Even when no disease-modifying drug exists, the genotype often dictates the right symptomatic choice — and the wrong one to avoid (see Epilepsy and Inborn Errors of Metabolism). The logic is mechanistic: each example below treats or avoids based on the specific molecular defect, not the seizure semiology. GLUT1 deficiency (SLC2A1) responds to the ketogenic diet because ketones bypass the defective glucose transporter to fuel the brain; pyridoxine-dependent epilepsy (ALDH7A1) responds to pyridoxine (B6) because supraphysiologic B6 overcomes the trapped-cofactor block. The genotype also flags drugs to avoid: in SCN1A/Dravet, avoid sodium-channel blockers (oxcarbazepine, lamotrigine), which worsen seizures because SCN1A loss already impairs inhibitory interneuron firing and further sodium-channel blockade compounds it; in SLC6A1, valproate is first-line (do not confuse with SCN1A); in ALDH5A1/SSADH, avoid vigabatrin (paradoxical worsening from excess GABA in a pathway already overloaded). The molecular diagnosis thus unlocks targeted therapy, flags contraindicated drugs, and ends the diagnostic odyssey.

Surveillance: the diagnosis triggers a program. A confirmed genetic diagnosis activates a defined surveillance schedule for known complications — a core reason genetic diagnosis changes management. The premise is that many genetic conditions carry predictable, clinically silent complications (a growing renal angiomyolipoma, an enlarging SEGA) whose outcomes improve markedly with early detection — so scheduled imaging substitutes for waiting on symptoms. Tuberous sclerosis complex (TSC) is the model, with clear international consensus (Northrup et al. 2021); see Neurodevelopmental Disorders for TSC clinical detail.

Bottom line: after a diagnosis, ask three questions — is there a disease-modifying therapy, does the genotype change symptomatic drug choices (including drugs to avoid), and what surveillance does this diagnosis now mandate?

A molecular diagnosis is rarely about one patient — it reshapes risk for the whole family. The full mechanics of these options live in Genetic Counseling; this section is the decision layer.

Start by naming the recurrence-risk scenario. The same diagnosis carries very different reproductive risk depending on how the variant arose:

• Inherited (one or both parents carry it): risk follows the inheritance pattern — 50% for autosomal dominant, 25% for autosomal recessive, and modified by penetrance and expressivity.
• Apparently de novo (absent in both parents' blood): empiric recurrence risk to a future pregnancy is low but NOT zero, because of possible parental germline (gonadal) mosaicism — commonly cited around ~1% for many neurodevelopmental conditions, though it varies by disorder (e.g., ~0.5–1% for MECP2/Rett, but ~7–14% for DMD). The mechanism: the variant arose post-zygotically in a clinically normal parent and seeded a subset of their germ cells, so it is enriched in gametes yet undetectable in the blood we test — which is exactly why a normal parental blood result cannot drive risk to zero. The disorder-specific spread (~1% vs. up to ~14% in DMD) reflects how early in germline development the typical mutation occurs. This is detailed in Somatic Mosaicism. Never counsel 'zero risk' from negative parental blood testing alone.

Cascade testing. Once a familial pathogenic variant is identified, single-site testing of relatives becomes cheap, fast, and unambiguous — the interpretive work is already done, so a relative's result is a clean yes/no rather than a fresh hunt. Counseling nuances follow from autonomy and the limits of actionability: testing minors for adult-onset conditions without childhood actionability is generally deferred until the child can consent for themselves (it forecloses their future choice and offers no childhood benefit); respect the right not to know; and obtain consent rather than assuming relatives want the result.

Reproductive options when the family variant is known:

Tie the option to the biology. Inheritance pattern and penetrance frame how a relative weighs these paths — a near-fully-penetrant dominant condition reads very differently from a low-penetrance or recessive one. The counselor's role is non-directive: present balanced options, not a steer.

A gene name is rarely a precise prognosis. The same molecular diagnosis can produce dramatically different trajectories, and four forces drive that uncertainty: variable expressivity (the same variant is mild in one person and severe in another, shaped by modifier genes, environment, and chance), incomplete penetrance (some carriers stay unaffected, so 'has the variant' is not 'has the disease'), small-n genotypes (only a handful of cases ever reported, so the apparent phenotype is a few anecdotes, not a distribution), and ascertainment bias — the earliest published cohorts skew toward the most severely affected because severe cases get tested and written up first, so the literature systematically reads worse than reality until milder cases surface. Recognizing this last force is what lets you honestly tell a family that the textbook picture is likely the dark end of a spectrum. Naming these forces is itself honest counseling — the principles of disclosure and non-directiveness live in Genetic Counseling and are applied to real families in Virtual Cases.

Heuristics for the prognosis conversation. Anchor explicitly on what is known versus unknown. Avoid false precision: give developmental ranges and the next milestones to watch rather than hard ceilings. Revisit the prognosis as the phenotype evolves — early prognosis is provisional. For an ultra-rare genotype, the honest answer is often 'we will learn together,' which builds more trust than a confident guess.

How to phrase it:

The diagnosis starts a coordinated program. A genetic result is the beginning of a multidisciplinary plan, not the end of one — because most neurogenetic conditions are multisystem, the diagnosis predictably generates parallel needs across organ systems that no single specialist owns. A typical team pairs clinical genetics and counseling with neurology, the therapies (PT, OT, speech-language pathology), and the organ specialists the syndrome demands — cardiology, nephrology, ophthalmology — plus developmental-behavioral support, palliative care where appropriate, and the family as a core partner. Left uncoordinated, this fragments into conflicting advice and dropped surveillance, so one clinician should serve as quarterback in a medical-home model: owning the problem list, synchronizing the specialists, and ensuring the surveillance schedule actually happens. Plan early for the pediatric-to-adult transition, a known point of dropout where established surveillance is most likely to lapse. The goal is a single coordinated plan, not a stack of disconnected referrals.