Large-Scale Cotton Genotyping: Choosing WGS, GBS, or SNP Panels for 1,000–10,000+ Samples

Figure 1: Cotton cohort planning connects biological sampling, platform choice and comparable data delivery; sample-count tiers are planning scenarios, not capacity promises.

Large-scale cotton genotyping succeeds when the selected loci, sampled material and production batches support the same research question. This guide explains how to choose WGS, restriction-based GBS or fixed SNP panels for 1,000–10,000+ samples, then specify pilot acceptance, cotton-specific QC and reusable deliverables.

Key takeaways

• Choose discovery breadth or repeatable fixed loci before committing the entire cohort to a platform.
• Distinguish conventional SNP microarrays from fixed-content GBTS liquid-targeted sequencing panels.
• Evaluate retained, informative At/Dt-assigned markers in the actual cotton material, not nominal panel density alone.
• Use representative pilots, balanced production batches and repeated bridge samples to assess comparability.
• Freeze sample identifiers, reference versions, locus definitions and filtering rules before combining data.

Define the Cotton Cohort

Start with the biological unit behind each barcode. A breeding line, a single segregating plant, an accession and a pooled seed lot are not interchangeable observations. Record whether DNA represents one plant or a deliberately defined pool, and whether field phenotypes refer to that same individual or its line. Technical repeats estimate process agreement; they do not add independent breeding material.

Separate Gossypium hirsutum, G. barbadense, interspecific derivatives and any diploid material in the manifest. A panel developed around elite upland cotton may not retain equally informative markers in all these groups. Likewise, a reference appropriate for one species does not automatically resolve divergent or introgressed segments in another. The existing cotton genome sequencing resource provides the broader genome context; a cohort brief needs the specific material actually being submitted.

Before counting samples, describe:

• biological groups, families, generations and expected relatedness;
• accession provenance, seed source and regeneration history;
• the individual or line matched to each phenotype record;
• expected within-line segregation and whether pooling is intentional;
• material that can be recollected if identity or DNA QC fails.

Keep these fields separate from plate location. Biological labels should survive re-extraction, resequencing and future submission rounds without changing the sample's permanent identity.

Choose Loci Before Platforms

The operational distinction is not simply sequencing versus arrays. Conventional SNP arrays read predefined sites through probe-based assays; fixed-content genotyping-by-target-sequencing, or GBTS, reads selected genomic targets through sequencing. Restriction-based GBS samples a reduced representation determined by the protocol and genome. WGS samples more broadly, but useful calls still depend on coverage, alignment and filtering.

Fixed loci and recurrent cohorts

The Cotton120K genotyping service currently describes a GBTS liquid-targeted sequencing panel, despite “array” in the page name. Specify its actual assay chemistry and manifest version in the quotation. Fixed content can simplify repeated comparisons, but it does not guarantee that every nominal locus passes QC or segregates in every family. Confirm the usable marker set on representative cotton samples before expansion.

Discovery and broader variation

Restriction-based GBS requires an agreed enzyme/library design and a strategy for missing or unevenly observed loci. Whole-genome resequencing is a route to broader SNP discovery, not a promise that every variant class will be recovered. Liu and colleagues used long-read pedigree sequencing and a graph-based analysis to investigate cotton structural variation [3]; that design should not be equated with routine short-read SNP genotyping.

A discovery subset followed by cohort-wide fixed-locus screening may be worth evaluating when both novel alleles and repeatable operational markers matter. Its size and composition should follow the diversity being represented. For a general technology comparison, use the LC-WGS, GBS and SNP-array selection resource; the cotton decision turns on whether the proposed calls remain usable across this cohort.

Plan Beyond One Thousand Samples

Scale changes coordination before it changes biology. More plates mean more opportunities for mislabeled wells, uneven arrival conditions, chemistry changes and incompatible delivery rounds. The following tiers illustrate planning priorities; they are not minimum sample requirements, validated provider throughput or turnaround commitments.

Decide who may authorize expansion and who resolves discrepant labels. A laboratory can finish a plate before the phenotyping team has reconciled accession identifiers. Without a shared release gate, apparently complete datasets can contain samples that cannot be joined to traits. Budget time for the biological owner to review exception reports, not just for library preparation and sequencing.

Sample number alone does not establish GWAS power or genomic-prediction accuracy. Gowda and colleagues combined 348 upland cotton accessions with evaluation across ten environments [2]. Li and colleagues examined 1,385 breeding lines while considering pedigree and environmental information [1]. Those studies illustrate the importance of phenotype design and validation; their cohort sizes are not universal thresholds for a different trait, population or program.

Count Informative Cotton Markers

For allotetraploid cotton, a locus needs a defensible genomic assignment. Similar At and Dt homeologous sequences can complicate mapping or assay specificity. Reads from two loci may produce apparent heterozygosity that is not variation at one correctly assigned locus. Conversely, filtering every heterozygous call would discard real segregation in appropriate populations.

Figure 2: Cotton QC distinguishes locus assignment from biological genotype variation; the diagram is conceptual rather than an experimental accuracy result.

Ask for evidence at the locus and group levels: mapping or probe specificity, allele balance where applicable, missingness, repeat agreement and segregation behavior. Genotype-model assumptions should fit the assay and material. A subgenome-resolved biallelic locus is not automatically a four-copy dosage marker merely because the organism is tetraploid. The polyploid crop genotyping resource gives additional context for assay transfer and locus specificity.

Report nominal targets, successfully assayed loci, QC-retained loci and informative polymorphic loci separately. A marker monomorphic within one family may still be informative across a diversity panel. Evaluate missingness and informativeness by species, family and introgression group rather than hiding differences in one cohort average. Assess whether retained markers adequately represent both subgenomes and the genomic regions important to the stated question.

Reference choice also affects the marker set. Feng and colleagues compared cotton references in a SLAF reduced-representation study [4]. Their results motivate explicit reference provenance, not an assumption that SLAF and every GBS protocol behave identically. Similar mapping percentages alone are insufficient evidence that two references produce interchangeable variant matrices.

Keep Tissue and DNA Traceable

At large scale, extraction responsibility should be settled before shipping. Starting from tissue gives the receiving laboratory control over extraction consistency, while submitted DNA reduces tissue handling but inherits variation from collection sites and extraction methods. Neither route is automatically preferable. Representative problematic samples should enter the pilot rather than being discovered after thousands of extractions.

Agree preservation and shipping conditions for the selected tissue and assay. Record collection dates, storage history and extraction method where known. Cotton preparations may require attention to inhibitors as well as DNA yield; concentration alone does not establish library compatibility. Confirm the current input mass, volume, concentration and integrity requirements with the selected service rather than importing a number from another platform's protocol.

• Use a permanent sample ID and a separate tube or well identifier.
• Include species, family, generation, source and submission-wave fields.
• Distinguish single-plant samples from intentional pools; document every pool's composition.
• Retain reserve tissue or DNA when feasible and record who holds it.
• Define how failed intake QC, recollection and re-extraction affect the manifest.

For segregating material, do not pool plants merely to achieve the requested DNA amount if individual genotypes are required. Archive a corrected manifest with both the original label and the reason for change. A silent relabeling can break phenotype joins even when the sequencing data themselves are technically acceptable.

Use a Representative Pilot

A convenient pilot is often an unrepresentative pilot. Include important cotton groups, close relatives, divergent accessions, introgression material and realistic variation in DNA preparation. Do not choose only the cleanest elite lines. Pilot size should follow the number of technical and biological risks being tested, with enough repeats to estimate agreement meaningfully.

Biological suitability of the marker set

Review retained-marker distribution, polymorphism and missingness within the intended populations. Ask whether the panel separates the relevant relationships without interpreting every cluster as a biological population. For a discovery route, inspect the shared-locus set and sensitivity to filtering. Where imputation is proposed, evaluate withheld-genotype agreement within representative groups and distinguish observed from imputed calls.

Operational readiness for production

The pilot should exercise intake, plate mapping, library preparation, analysis and delivery formats, not just generate a successful sequencing run. Set acceptance criteria before reviewing the full biological pattern. Criteria can address sample pass/fail logic, repeat concordance, retained loci, ambiguity handling and compatibility with legacy samples; numerical cutoffs must be agreed for the assay and intended analysis.

Record the decision as proceed, revise or hold. Preserve the report, accepted manifest/reference versions and unresolved limitations. A revised extraction method or panel should trigger the relevant repeat checks before production resumes. This record establishes what was actually validated, rather than allowing “pilot passed” to become an undefined claim.

Bridge Production Batches

Avoid assigning every breeding group or collection year to its own laboratory batch. Balance biological groups across plates and submission waves where collection constraints permit. Keep randomization reproducible by retaining the allocation record. When staged arrival prevents full balancing, acknowledge the confounding and use shared samples to assess comparability across waves.

Figure 3: Bridge samples connect production waves, while blinded repeats test agreement within the defined process; they do not increase biological sample size.

A bridge is the same biological material or DNA tested across batches. An independently collected plant from the same accession is not an equivalent technical control when the accession can segregate. Extraction repeats test a broader process than repeated libraries from one DNA aliquot. Define which process each control is intended to evaluate and maintain sufficient reserve material.

Review discordance by locus, sample and batch, separating callable-site genotype disagreement from differences in missingness. Include non-reference and heterozygous comparisons where appropriate; an overall agreement value dominated by common homozygous calls can hide useful discrepancies. Set batch-release decisions against the predefined criteria and record whether a failed sample, locus or entire batch requires rework. Post hoc batch correction cannot reconstruct lost genotypes or remove identity uncertainty.

Harmonize Before Merging Genotypes

Two VCF files are not compatible simply because their chromosome columns look similar. Lock the reference assembly and its release, chromosome naming, coordinates, REF/ALT definitions and locus identifiers. Normalize variants consistently and preserve the original assay target or probe ID. Record the mapping used when translating legacy array alleles to a reference-oriented representation.

For panel, GBS and WGS datasets, start with verified overlapping loci. Establish equivalent alleles and orientation before evaluating concordance on shared samples. Do not substitute missing calls with the reference allele, or treat every absent site as a confident homozygous genotype. Separate multiallelic sites, ambiguous mappings and sites requiring remapping from the clean intersection.

• Freeze assay manifests, reference files, caller versions and filtering parameters.
• Keep raw observations and filtered genotypes as distinct data layers.
• Document changes in retained loci and failed samples between releases.
• Validate imputation for the relevant cotton groups and retain its provenance.
• Preserve platform-specific QC even when a common analysis matrix is delivered.

A panel update may improve coverage while reducing direct compatibility with earlier rounds. Agree whether analysis will use the common intersection, separately modeled datasets or a validated imputation strategy. Coordinate translation alone is not validation: a correctly converted position can still represent an ambiguous or differently assayed locus.

Specify Reusable Data Deliverables

Make delivery requirements part of the quotation rather than an afterthought. An acceptance package should let the research team reproduce which samples and loci entered each analysis. File availability depends on the assay: sequencing generates reads, whereas a conventional SNP array generates platform-specific intensity and genotype data, not FASTQ files.

Figure 4: Harmonized delivery retains provenance and QC alongside genotypes; optional trait analyses require separately matched phenotype data.

Genomic selection is a separate modeling task, not an automatic output of high-throughput genotyping. Billings and colleagues evaluated public cotton breeding material using genotype data and repeated field trials [5]. Define training/test relationships and appropriate family, year or environment holdouts before claiming predictive usefulness. Likewise, diversity patterns and associated loci remain conditional on population sampling, marker coverage and analysis assumptions; none constitutes guaranteed breeding gain.

Prepare the Cotton Project Brief

Decide which work remains in-house: collection, extraction, phenotype curation, analysis or long-term database maintenance. Outsourcing assay production does not transfer responsibility for biological identities or field design. Assign owners for each handoff, exception decision and release approval. Consider the total scope of extraction, pilot work, rework, sequencing or assay intensity, analysis and storage when discussing cost and schedule; no universal per-sample price or delivery time applies.

For a cotton genome sequencing project discussion, supply species and material groups, anticipated cohort and arrival waves, tissue/DNA availability, discovery versus fixed-locus objectives, legacy genotype files, required variant classes and phenotype plans. Request an agreed pilot, production controls, harmonization rules and delivery specification. That brief makes a proposed WGS, GBS or SNP-panel route assessable against the cotton research question rather than against sample count alone.

Cotton Cohort Genotyping FAQ

References

1. Li Z, Liu S, Conaty W, et al. Genomic prediction of cotton fibre quality and yield traits using Bayesian regression methods. Heredity. 2022;129(2):103–112. doi:10.1038/s41437-022-00537-x.
2. Gowda SA, Fang H, Tyagi P, et al. Genome-wide association study of fiber quality traits in US upland cotton (Gossypium hirsutum L.). Theoretical and Applied Genetics. 2024;137(9):214. doi:10.1007/s00122-024-04717-7.
3. Liu S, Zuo D, Cheng H, et al. Cotton pedigree genome reveals restriction of cultivar-driven strategy in cotton breeding. Genome Biology. 2023;24(1):282. doi:10.1186/s13059-023-03124-3.
4. Feng Z, Li L, Tang M, et al. Detection of Stable Elite Haplotypes and Potential Candidate Genes of Boll Weight Across Multiple Environments via GWAS in Upland Cotton. Frontiers in Plant Science. 2022;13:929168. doi:10.3389/fpls.2022.929168.
5. Billings GT, Jones MA, Rustgi S, et al. Outlook for Implementation of Genomics-Based Selection in Public Cotton Breeding Programs. Plants. 2022;11(11):1446. doi:10.3390/plants11111446.

This content and the described services are intended for agricultural and biological research. CD Genomics does not offer these services for clinical diagnosis, treatment decisions or individual health assessment.