Sorghum Genotyping for Diverse Germplasm Collections: 20K Panel, GBS, or WGS?

Figure 1: Choose between a known marker backbone and broader variant discovery according to the collection's biological composition and research purpose.

A fixed 20K panel can provide a consistent marker backbone for sorghum germplasm characterization when its loci are informative across the collection. GBS adds discovery within reduced genomic regions; WGS expands discovery across more of the genome. The decision depends on which accessions and alleles must be represented, not nominal marker count alone. A representative pilot and separate conservation and GWAS datasets make that choice easier to defend.

Key takeaways

• Define whether each genotype represents a plant, a line or a heterogeneous accession before choosing an assay.
• Check usable, polymorphic markers within underrepresented sorghum groups, not just across the whole collection.
• Keep rare-allele conservation separate from the marker filtering used for structure analysis and GWAS.
• A representative core collection is not automatically a sufficiently powered association panel.

Define the Sorghum Sampling Unit

A collection of named accessions is not necessarily a collection of genetically uniform lines. An accession may be an inbred breeding line, a landrace maintained as a seed population, or material regenerated several times since collection. Genotyping one plant describes that plant. It does not establish the full diversity of a heterogeneous seed lot.

Decide what the project needs to resolve: relationships among representative plants, variation within accessions, or identity across regeneration lots. Pooling tissue changes the biological measurement and should not be treated as an ordinary individual genotype without an agreed calling model. Sampling additional plants within selected accessions may address a conservation question that increasing marker density alone cannot answer.

• Connect every extraction to accession ID, individual plant ID and regeneration or seed-lot record.
• Retain collection location, botanical race where available, improvement status and known pedigree or introgression history.
• Flag uncertain species identity, suspected mixtures and material requiring a different ploidy or genotype-calling model.
• Keep the sampled plant traceable to the material actually used in phenotyping or crossing.

Resolve these distinctions before samples leave the genebank. Otherwise, a precise genotype table can answer a different question from the one the accession register was intended to support.

Separate Conservation From Trait Mapping

For conservation, the objective may be to represent genetic groups and retain uncommon alleles in a manageable core or mini-core. For GWAS, the objective is to test genotype–trait relationships with adequate carriers, credible phenotypes and manageable confounding. The same accession subset can serve both purposes, but that must be demonstrated rather than assumed.

Mufumbo and colleagues assembled a 310-accession core from 3,333 Ugandan sorghum accessions using DArTseq-derived markers [1]. This is evidence that genomic data can support core selection, not a rule that every collection should use the same fraction or selection settings.

For a mini-core, document what was preserved and what was sacrificed. Compare subgroup representation and allele retention with the original collection, then retain explicit exceptions for important traits or provenance. A genetically distant accession may be valuable, but distance alone does not make it the best breeding parent.

Figure 2: Validate a core against the full collection; genetic representation and phenotypic representation require separate evidence.

Use a Known Marker Backbone

Fixed content supports a common coordinate set

A fixed panel is attractive when the main need is to compare many accessions at predefined loci and revisit those loci in future additions. The sorghum 20K genotyping panel provides a crop-specific option for this known-marker approach. Its nominal target count is a starting specification, not the number of polymorphic, high-confidence loci that every collection will retain.

A marker can be successfully called yet monomorphic in a particular subgroup. Conversely, a locus informative in one breeding pool may be difficult to call in divergent material. Evaluate both technical callability and biological informativeness. Reporting only a collection-wide average can conceal a poorly represented geographic or racial group.

Inspect the assay definition before choosing

Request the current assay chemistry, panel version, locus manifest, reference assembly and coordinate convention. “20K” describes nominal marker content; it does not identify whether a product uses a conventional array or a sequencing-based targeted assay. The fixed-array and GBTS chemistry comparison explains that distinction without determining the chemistry of a particular sorghum offering.

Pan-genome-informed marker selection can broaden the discovery background, but it does not remove every ascertainment effect. Ask which populations informed marker selection and inspect allele frequencies in your own material. A fixed panel is reasonable when the known backbone answers the question; it is less suitable when finding variation absent from that backbone is the central objective.

Discover Beyond the Fixed Panel

GBS samples a reduced genomic fraction

Restriction-based GBS can discover SNPs in sampled regions instead of restricting analysis to a predefined SNP list. Its useful marker set depends on the protocol, restriction sites, library representation, sequencing allocation and analysis filters. These dependencies matter when the collection combines deeply separated groups or when future additions must be compared with an existing dataset.

Use an agreed protocol and analysis version for comparability. A missing GBS observation is not automatically a reference allele; restriction-site differences and insufficient observations can both contribute. DArTseq studies provide relevant reduced-representation examples, but DArTseq is not interchangeable with every restriction-based GBS protocol or the CD Genomics fixed 20K panel.

WGS expands the discovery space

Whole-genome resequencing is a stronger candidate when broader SNP and indel discovery, sequence reuse or detailed investigation of divergent material is required. Ruperao and colleagues analyzed genomic resources from 272 sorghum accessions across races [4]. Habyarimana and colleagues examined cultivated sorghum and recombinant S. bicolor × S. halepense lines [5], illustrating why the biological composition of a discovery set matters.

WGS still has limits: coverage, alignment, reference representation and variant-calling assumptions determine what is observable. If structural variation is a primary endpoint, specify the relevant variant classes, sequencing technology and validation plan. Routine short-read resequencing should not be described as detecting every structural change.

Test Underrepresented Sorghum Groups

A convenient pilot made entirely from well-characterized elite lines does not qualify an assay for a geographically diverse landrace collection. Include material representing the groups most likely to challenge marker informativeness or sequence alignment, as well as the routine samples expected in production.

Select pilot entries across available racial, geographic and improvement-status categories. Include unusual introgression material and different DNA-source or storage categories where these are in scope. Botanical race and geography are useful selection variables, not substitutes for genetic grouping. The pilot size should reflect heterogeneity and the decisions it must support, not an arbitrary fraction of the final cohort.

• Summarize sample and locus missingness within each represented group.
• Separate monomorphic loci from failed observations and report retained markers on each chromosome.
• Check a suitable repeat or orthogonal subset at shared loci, distinguishing discordance from missing calls.
• Review unexpected heterozygosity, identity discrepancies and unusual mapping behavior before scaling.

Agree how failed categories will be handled: revised extraction, additional observations, alternative assay or a separately analyzed subgroup. Set acceptance criteria before evaluating results, while allowing documented changes when the pilot reveals a biological assumption was wrong. A good pilot produces a defensible inclusion rule and exception list, not just a favorable mean call rate.

For example, if a small landrace group has unusually high missingness, first compare extraction quality and observed read or assay evidence with the other groups. If technical quality is comparable but particular loci repeatedly fail in that group, investigate locus representation or reference compatibility. Excluding the whole group would make the dataset look cleaner while weakening the collection's conservation value. Document that tradeoff explicitly if exclusion becomes unavoidable.

Preserve Rare and Group-Specific Alleles

One filtered SNP matrix should not be expected to support every purpose. A global minor-allele-frequency filter can remove variants concentrated in a small geographic group, even when those variants are relevant to conservation. Low-frequency calls also need evidence of reliability; preserving them does not mean accepting every isolated observation.

Keep a quality-controlled conservation archive, a structure-analysis marker set and trait-specific association sets. The archive retains supported uncommon variation and its uncertainty. The structure set may use LD pruning so dense regions do not dominate relationships. The association set needs appropriate allele-count and quality rules within the individuals phenotyped for that trait.

Kasule and colleagues reported 20,211 initial DArTseq SNPs and retained 7,156 after their study's filters [2]. These numbers belong to that dataset, not the CD Genomics 20K panel, and are not recommended universal thresholds. They show why initial marker count and the analyzed set must be reported separately.

Evaluate allele retention when selecting a core, not only mean diversity statistics. A subset can resemble the full collection on averages while losing an uncommon subgroup. The GBS diversity and core-collection resource provides supporting analysis context; for sorghum, the conservation objective should determine which filtered view is used for selection.

Figure 3: Preserve one documented genotype archive while deriving different marker sets for conservation, population structure and trait association.

Read Structure Alongside Passport Records

PCA, kinship and ancestry analyses can expose relationships that passport labels miss. They can also reflect uneven sampling, correlated missingness or strongly represented genomic regions. Inspect these factors before interpreting a cluster as a biologically distinct group. Compare pruned and alternative reasonable marker sets when the interpretation will affect accession selection.

In their study of Ethiopian landraces, Girma and colleagues found genetic groupings that only partly overlapped racial and geographic descriptors [3]. That supports using both genomic and passport evidence. It does not justify discarding passport records when the two disagree.

Genetic diversity assessment can be scoped around this combined interpretation: allele frequencies, diversity within groups, between-group relationships and accession-level exceptions. Request tables behind plots so curators can inspect the observations associated with each accession.

Close genetic similarity identifies candidates for identity review, not automatic deletion. Different names can describe related material; separate accessions can also carry different provenance, cultural value or unmeasured traits. Confirm suspected duplicates with source records and additional sampling where appropriate. Likewise, genetic distance may guide parent exploration, but does not establish heterosis, combining ability or performance in a target environment.

Qualify the Phenotyped GWAS Subset

A collection-wide genotype matrix is only the first input to GWAS. Each trait usually has a smaller matched subset because some accessions were not grown, failed establishment or lack acceptable measurements. Recalculate allele frequencies, missingness, kinship and structure in that subset. Statistics from the full genebank can misrepresent the actual association analysis.

Do not assume 20K is sufficient or insufficient for every sorghum GWAS. Relevant density depends on LD, population composition and the trait architecture. More markers cannot compensate for a rare allele carried by too few phenotyped individuals. A mini-core optimized for diversity may deliberately include unusual accessions without enough carriers for robust trait testing.

Inspect coverage across the genome rather than treating evenly numbered chromosome markers as evenly informative evidence. LD can differ among groups, and a single pooled decay estimate may hide those differences. Where a subgroup cannot support reliable estimation, report that limitation instead of supplying a confident density recommendation. The relevant conclusion may be that additional phenotyping or population development is needed before changing the genotyping technology.

For agricultural GWAS analysis, supply trial design and phenotype-processing decisions alongside genotypes. Models and sensitivity checks should address kinship, population structure and environmental effects without claiming to eliminate all confounding. The GWAS interpretation and limitations guide explains why association signals remain candidate evidence rather than causal proof.

Deliver Separate Sorghum Evidence Packages

Request a shared genotype foundation and two purpose-specific analysis packages. The collection package supports curation and conservation decisions. The trait package supports specified phenotype analyses. Separating them prevents a stringent GWAS filter from silently becoming the conservation definition of diversity.

• Shared foundation: sample manifest, assay and reference versions, locus definitions, genotype observations, QC metrics and an exclusion ledger.
• Collection package: group-aware allele summaries, diversity and relationship tables, core-selection criteria, retained-allele checks and curator exceptions.
• Trait package: phenotype matching, trait-specific sample and marker counts, covariate definitions, LD and kinship summaries, association outputs and sensitivity notes.

Record whether calls were observed or imputed and which input files and reference population support imputation. An imputed value is not equivalent to direct observation for every use. Keep original observations available so later conservation questions are not forced through a trait-mapping filter.

If combining fixed-panel and sequencing data, harmonize only documented compatible loci. Resolve assembly versions, allele orientation and variant representation; a nearby coordinate is not sufficient evidence of identity. Compare shared-locus concordance on suitable samples before treating the combined matrix as one dataset. Use discovery variants absent from the panel in a separately labeled analysis rather than implying all platforms measured them.

Figure 4: A versioned genotype foundation supports distinct collection-management and trait-mapping packages with their own inclusion rules.

Scope the Sorghum Genotyping Decision

Start the project discussion with the accession inventory, sampling unit, underrepresented groups, available DNA, phenotype coverage and intended use of the final collection. Then agree the fixed-marker or discovery requirement, pilot acceptance questions and analysis deliverables. These inputs are more useful than asking for the largest possible SNP count.

CD Genomics can help scope the sorghum panel, GBS or resequencing route alongside diversity assessment or GWAS needs. A known-marker backbone is appropriate when validated informativeness meets the collection objective; discovery is preferable when missing biological variation is the question. Preserve the distinction in the statement of work so a technically successful assay also delivers the evidence the genebank or breeding team needs.

Sorghum Germplasm Genotyping FAQ

References

1. Mufumbo R, Chakrabarty S, Nyine M, et al. Genomics-based assembly of a sorghum bicolor (L.) moench core collection in the Uganda national genebank as a genetic resource for sustainable sorghum breeding. Genetic Resources and Crop Evolution. 2023;70(5):1439–1454. doi:10.1007/s10722-022-01513-4.
2. Kasule F, Alladassi BME, Aru CJ, et al. Genetic diversity, population structure, and a genome-wide association study of sorghum lines assembled for breeding in Uganda. Frontiers in Plant Science. 2024;15:1458179. doi:10.3389/fpls.2024.1458179.
3. Girma G, Tirfessa A, Bejiga T, et al. Assessing genetic, racial, and geographic diversity among Ethiopian sorghum landraces and implications for heterotic potential for hybrid sorghum breeding. Molecular Breeding. 2024;44(7):46. doi:10.1007/s11032-024-01483-8.
4. Ruperao P, Gandham P, Odeny DA, et al. Exploring the sorghum race level diversity utilizing 272 sorghum accessions genomic resources. Frontiers in Plant Science. 2023;14:1143512. doi:10.3389/fpls.2023.1143512.
5. Habyarimana E, Gorthy S, Baloch FS, et al. Whole-genome resequencing of Sorghum bicolor and S. bicolor × S. halepense lines provides new insights for improving plant agroecological characteristics. Scientific Reports. 2022;12(1):5556. doi:10.1038/s41598-022-09433-0.

CD Genomics services described here are for agricultural research use only, not clinical diagnosis or patient testing.