Custom NGS Probe Design: Inputs, Design Checks, and Capture Validation

A custom NGS capture project starts with three decisions: which sequences must be interrogated, what material will enter the workflow, and what evidence will show that the assay meets its research objective. A gene list alone does not resolve these questions.

Before approving a probe design, define the target coordinates, identify difficult regions, confirm compatibility with the intended library preparation and capture workflow, and establish a pilot evaluation plan. Keep design coverage separate from experimental coverage: including a region in a design does not establish that every base will receive sufficient usable sequencing depth.

Translate the research question into sequence requirements

Specify what the experiment needs to detect or measure. Coding variants, splice-site variants, genomic rearrangements, expressed fusions, and transcript abundance require different target definitions.

For a DNA project, identify whether the requested regions include coding exons only, exon boundaries, untranslated regions, selected introns, or other genomic intervals. For an RNA-derived project, identify the transcripts, exons, isoforms, or fusion events of interest.

Separate essential regions from optional additions. This gives the design team a basis for discussing exclusions and trade-offs rather than treating every interval as equally important.

Fix the reference and coordinate conventions

Provide the species, reference assembly, and annotation release used to generate the targets. If the starting point is a gene list, specify the transcript selection policy: one selected transcript per gene, a defined transcript set, or all relevant annotated transcripts.

The project documentation should identify:

·       Reference assembly and chromosome naming convention.

·       Target file format and coordinate convention.

·       Transcript identifiers and versions, where relevant.

·       Whether exon flanks or other padding are already included.

·       Regions that must remain under consideration even if they are difficult to target.

Retain the approved target file as a versioned project document. Design review and sequencing evaluation should refer to the same target definition unless a change is explicitly documented.

Distinguish requested targets from probe locations

The regions you want to analyze are not necessarily identical to the regions where probes are placed. Probes may overlap target boundaries or bind nearby sequences on library fragments that also contain the target.

This distinction matters during evaluation. Broad Institute’s CollectHsMetrics documentation requires separate bait and target interval files. Bait intervals describe probe-associated regions; target intervals define the regions whose coverage is being evaluated.

Request a clear record of the original targets, final evaluation targets, and exclusions. Difficult regions should not disappear from the coverage denominator without an explicit decision.

Probe count and nominal tiling density describe aspects of a design, but they do not establish capture performance. A useful design review explains where enrichment may be difficult and how those risks will be investigated.

Design consideration	What to review before approval	What to examine in the pilot
Target boundaries and short intervals	Probe placement relative to essential bases and the expected library insert distribution	Coverage across the full interval, including its edges
GC composition	Regions with extreme GC composition and proposed design adjustments	Coverage patterns across different GC ranges
Repeats and low-complexity sequence	Masked regions, retained exceptions, and possible alternative placements	Off-target enrichment and persistent coverage gaps
Homologous genes and pseudogenes	Potential cross-hybridization and limitations on unique read alignment	Mapping quality and locus-specific interpretability
Probe representation	Whether representation is uniform or adjusted, and why	Overrepresented and underrepresented intervals

Treat GC composition as a risk indicator

GC composition can affect the coverage obtained from enrichment and sequencing workflows, but a fixed acceptable GC range is not a universal design rule.

In a comparison of genomic enrichment methods, bases without genotype assignments were more GC-rich than the overall region of interest. This supports checking for GC-associated dropout; it does not establish a guaranteed cutoff for another platform or workflow.

Ask how difficult GC regions will be handled and how their performance will be reported. Do not assume that additional probes or deeper sequencing will resolve every gap.

Separate capture specificity from mapping specificity

A probe may hybridize to homologous sequences, while reads from an enriched fragment may be difficult to assign uniquely. These are related but distinct problems.

The foundational solution hybrid-selection study by Gnirke and colleagues describes avoiding repetitive sequences during bait design or blocking them during hybridization. These strategies address enrichment challenges, but they do not make every homologous locus uniquely interpretable.

Review both potential non-target binding and the alignment strategy for essential difficult regions. Request an exception list with a documented status for each: retained, redesigned, excluded with justification, or requiring separate evaluation.

“RNA project” describes the biological starting material, not necessarily the chemistry of the capture probe or the material entering hybridization. An RNA-derived sequencing workflow may convert RNA into cDNA and then enrich the resulting library. Cabanski and colleagues demonstrated this approach in their study of cDNA hybrid capture from low-input and archived samples.

For DNA projects, provide the genomic regions and variant classes of interest, together with the expected sample quality and fragment characteristics.

For RNA-derived projects, additionally define:

·       The transcript annotation and isoforms of interest.

·       Whether the objective is expression analysis, expressed-variant detection, fusion detection, or splice analysis.

·       Whether fusion detection is limited to specified events or includes discovery involving selected genes.

·       The library preparation method and molecular material entering capture.

·       The controls and analysis pipeline needed to evaluate the intended endpoint.

Do not apply a DNA coverage requirement unchanged to an RNA-derived library. Transcript abundance varies, and capture can alter representation. A published exome RNA sequencing study demonstrated enrichment of rare transcripts and splice-junction evidence while also discussing limitations relevant to quantitative expression analysis.

For an RNA project, evaluate whether the workflow supports the intended transcript-level measurement—not whether every targeted gene reaches comparable depth.

Use a staged workflow so that sequence-design decisions, product quality control, and assay performance remain separate.

1.     Approve the target definition. Resolve the reference assembly, annotation, padding, priorities, and exclusions.

2.     Review the proposed design. Confirm difficult-region handling, workflow compatibility, and the design information to be supplied.

3.     Agree on synthesis and release documentation. Establish which QC checks apply, what they measure, and which records will accompany delivery.

4.     Run pilot captures. Use material representative of the intended sample types and input conditions, with suitable controls.

5.     Sequence and analyze consistently. Record sequencing allocation, reference files, software versions, and filtering rules.

6.     Investigate failures before revision. Review recurring gaps, off-target enrichment, and duplicate burden before changing probe placement or representation.

Do not substitute synthesis QC for capture validation. Product release checks assess the supplied material; sequencing-based evaluation assesses its performance within the complete experimental workflow.

Document library preparation, adapter configuration, insert-size distribution, pooling, capture conditions, and amplification. Published capture studies, including the cDNA hybrid-capture study, evaluate enrichment within defined experimental workflows rather than treating probe design as the only determinant of the result.

A panel can have high mean depth while leaving essential intervals insufficiently covered. Evaluate target completeness, coverage distribution, enrichment, and duplication together.

Broad Institute’s Picard metric definitions distinguish enrichment metrics from filtered target-coverage metrics. Different metrics may use different subsets of the sequencing data, so apparently similar percentages are not necessarily comparable.

Metric	What it helps answer	What must accompany the result
Mean target coverage	How much filtered depth was obtained across the evaluation targets?	Target file, filtering rules, and duplicate handling
Fraction of target bases at or above a chosen depth	How completely did the sample meet the project’s depth requirement?	Depth threshold, denominator, and essential-region results
Coverage uniformity	Is usable depth distributed evenly or concentrated in a subset of targets?	Metric definition and interval-level coverage
On-target fraction	How much analyzed sequence overlaps the intended targets?	Whether reads or bases are counted, denominator, padding, and filters
Off-bait fraction	How much aligned sequence falls away from bait-associated regions?	Bait intervals and any near-bait definition
Duplicate rate	How much aligned sequence is classified as duplicate?	Duplicate-identification method and sequencing context

Picard separately defines on-target, on-bait, and near-bait quantities. A metric that includes near-bait sequence should not be presented as equivalent to sequence falling inside the evaluation targets.

Set acceptance criteria before reviewing the pilot results. Required depth and breadth should follow the research objective and analysis method; no single threshold establishes suitability for every custom panel.

Compare pilots on a consistent basis

Use the same evaluation targets and analysis settings when comparing designs or conditions. Report sequencing allocation as well as usable coverage. Where appropriate, compare results at matched read counts to distinguish enrichment differences from differences in sequencing effort.

Examine essential intervals individually. A satisfactory panel-wide average should not override a persistent gap in a must-cover region.

For RNA-derived libraries, include endpoint-specific evidence such as control detection, fusion-supporting reads, splice-junction evidence, or expression reproducibility. Genomic coverage metrics alone do not demonstrate that the intended transcript-level measurement works.

Before requesting a quotation or design assessment, assemble the information needed to define both the product and its evaluation:

·       Project objective and intended research endpoint.

·       Species, reference assembly, and annotation release.

·       Target interval file or gene and transcript list.

·       Must-cover regions, optional targets, and acceptable exclusions.

·       Sample types, preservation conditions, available input, and quality information.

·       Library preparation method, adapter configuration, and expected insert-size distribution.

·       Intended sequencing platform and read configuration.

·       Expected sample numbers and pooling arrangement.

·       Pilot controls, analysis pipeline, and proposed acceptance criteria.

·       Required design files, QC documentation, and revision arrangements.

Clarify which deliverables are included in design and synthesis, and which questions require experimental evaluation. A quotation for probe production should not be interpreted as evidence that a particular sample type or target set has already been validated.

For projects ready for a technical discussion, explore Dynegene’s NGS custom probe design and synthesis service. Provide the target files and workflow requirements above so that the discussion can address the relevant design, delivery, and evaluation needs.

Approve the design when the target definition is stable and its limitations are documented. Accept the assay when representative pilot data meet the predefined research criteria.