What Should a Population Genomics Report Include? Data Files, Figures, QC Metrics, and Interpretation
For Research Use Only. Not for use in diagnostic procedures or clinical decision-making.
Researchers who are new to population genomics outsourcing often expect to receive a VCF file and a few PCA plots. What they should receive is a structured, publication-grade report that documents every analytical decision, reports every relevant QC metric, and provides the data provenance needed to defend the work to reviewers. The gap between those two expectations — a data dump versus a deliverable — is where project value is either created or destroyed.
This guide defines the minimum components of a credible population genomics report: raw and processed data files, variant calling outputs with QC tables, population structure visualizations, diversity and selection metrics, and a publication-ready methods summary with interpretation notes. Use this as a checklist when defining deliverables with a provider, and as a benchmark when reviewing what you receive.
Figure 1: The six layers of a complete population genomics report — each layer builds on the data layer below, and missing layers compromise the defensibility of the final conclusions.
What a Complete Report Contains
A population genomics report is not a single document — it is a package of data files, figures, tables, and text that together provide a complete, reproducible account of the project. A credible report can be read by a peer reviewer who has never seen your samples and has no access to your raw data, and that reviewer should be able to assess whether the analysis was performed correctly.
The six essential layers are:
- Raw data and processing outputs. The files that constitute the analytical foundation — FASTQ, BAM, or VCF — along with read-level and alignment-level QC metrics that demonstrate the sequencing and processing were technically sound.
- Variant calling and QC tables. The VCF file itself, plus variant-level QC statistics (depth distribution, missingness rate, transition/transversion ratio, heterozygosity distribution) that demonstrate the variant call set meets population genomics standards.
- Population structure visualizations. PCA, ADMIXTURE, and phylogenetic trees — the three core analyses that reveal how your samples relate to each other and whether population labels match genetic ancestry.
- Diversity and selection metrics. Within-population diversity (π, θW, HE), between-population differentiation (FST), LD decay, and selection scan results (iHS, XP-EHH, Tajima's D) that quantify the evolutionary forces shaping your study populations.
- Methods summary. A complete, version-numbered description of every software tool and non-default parameter used — the content you will paste directly into your manuscript's methods section.
- Interpretation notes. Plain-language explanations of what each figure and table shows, what the QC metrics mean for downstream use, and which results are robust versus which require cautious interpretation.
The following sections detail what each layer should contain. Before commissioning a project, use our population genomics project quote checklist to define your deliverable expectations. When comparing providers, our provider selection guide includes specific questions about reporting standards.
Raw Data and Processing Outputs
The report should begin with what was sequenced and what happened to the reads. This section is not glamorous, but it is the foundation of reproducibility — and the section that reviewers are most likely to scrutinize if your downstream results are unexpected.
Table 1: Raw Data and Processing Deliverables
| Deliverable | What It Should Include | Why Reviewers Check It |
| Per-sample raw read QC | Read count, Q30 percentage, GC content, adapter content, duplication rate | Low Q30 or high duplication indicate library preparation problems |
| Alignment summary | Total reads, mapped reads, mapping rate, properly paired rate, insert size distribution | Low mapping rate suggests reference genome mismatch or contamination |
| Coverage statistics | Mean coverage, percentage of genome at 5×/10×/20×, coverage distribution plot | Uneven coverage across samples indicates batch effects or library degradation |
| FASTQ files | Raw or trimmed reads; specify whether adapter trimming was performed | Required for re-analysis with improved reference genomes or pipelines |
| BAM files | Coordinate-sorted, duplicate-marked (not removed), indexed | Required for inspecting variant calls at specific loci |
For projects that started from existing FASTQ, BAM, or VCF files, the report should note at which pipeline stage the provider received the data and what processing steps they performed from that point forward.
Variant Calling and QC Tables
The variant call set is the core deliverable. It must be accompanied by QC metrics that demonstrate its reliability for population genomics analysis. A VCF file without QC context is a collection of unvalidated assertions.
The report should include:
- Per-sample variant call statistics. Total SNPs, total indels, heterozygous/homozygous ratio, and transition/transversion (Ti/Tv) ratio. Ti/Tv ratios for whole genome sequencing typically range from 2.0–2.3 for most vertebrate genomes; values outside this range suggest false positive variant calls or systematic sequencing errors.
- Missingness report. Per-sample missingness rate and per-variant missingness rate. Samples with >10% missingness should be flagged for review. Variants with >10% missingness may need to be excluded from certain analyses, or the missingness pattern should be investigated for batch effects.
- Depth distribution. Mean depth, depth standard deviation, and depth distribution histogram across all variant sites. Depth outliers — variants with extremely high coverage — may represent paralogous or repetitive regions where reads from multiple genomic locations map to a single reference position, producing false heterozygous calls.
- Hardy-Weinberg equilibrium (HWE) summary. For diploid organisms within reasonably panmictic populations, HWE p-values identify genotyping errors, population substructure, or selection. The report should state the HWE threshold used for filtering and the number of variants excluded.
- MAF spectrum. Minor allele frequency distribution across all variants. An excess of low-frequency variants (singletons, doubletons) may reflect sequencing error, while a deficit may indicate overly aggressive filtering or population demographic history.
The variant calling pipeline — software, version, filtering thresholds — should be documented in the methods summary so these QC metrics can be interpreted in the context of how they were generated.
Population Structure Visualizations
Population structure analysis answers the most fundamental question in population genomics: how are your samples related to each other, and do the genetic relationships match your sampling design?
A complete report includes three complementary structure visualizations:
- PCA plot. Principal component analysis of genotype data, typically showing PC1 vs. PC2 (and PC1 vs. PC3 in a supplementary figure) with samples color-coded by population label. PCA reveals whether populations are genetically distinct, whether there are outlier samples, and whether population labels correspond to genetic clusters. The report should state the number of variants used (after LD pruning), the LD pruning parameters (window size, step size, r² threshold), and the percentage of variance explained by each PC.
- ADMIXTURE analysis. Model-based ancestry estimation at multiple K values (number of assumed ancestral populations), displayed as a stacked bar plot. Cross-validation error should be reported for each K, with the recommended K indicated. ADMIXTURE reveals admixture, migration, and substructure that PCA alone may miss — particularly when populations have experienced gene flow.
- Phylogenetic tree. A neighbor-joining or maximum-likelihood tree based on genome-wide genetic distances, with bootstrap support values on major branches. The tree provides a hierarchical view of population relationships and is often the most intuitive visualization for communicating results to non-genomics collaborators.
These three analyses — PCA, ADMIXTURE, and phylogenetic trees — are complementary, not redundant. PCA is fast, model-free, and good at detecting outliers. ADMIXTURE explicitly models ancestry proportions and is better at detecting admixture. Phylogenetic trees provide bootstrap support values that quantify statistical confidence in population relationships. A population structure analysis that includes all three provides a comprehensive view.
Figure 2: Complementary population structure visualizations — PCA (left), ADMIXTURE (center), and phylogenetic tree (right) — each reveals different aspects of population relationships that the others may miss.
Diversity and Selection Metrics
Beyond structure, the report should quantify genetic diversity within populations, differentiation between populations, and evidence of natural selection. These metrics move the report from "what your data looks like" to "what biological processes shaped it."
Table 2: Core Diversity and Selection Metrics
| Metric | What It Measures | Standard Visualization |
| Nucleotide diversity (π) | Average pairwise nucleotide differences within a population | Per-population bar chart or sliding-window genome scan |
| Watterson's θW | Population mutation rate estimated from segregating sites | Compared with π — discrepancy signals demographic change or selection |
| Haplotype diversity (HE) | Expected heterozygosity under HWE | Per-population summary table |
| FST | Genetic differentiation between population pairs | Pairwise FST heatmap or per-site Manhattan plot |
| LD decay (r² vs. distance) | Linkage disequilibrium as a function of physical distance | LD decay curve, per-population or per-chromosome |
| Tajima's D | Deviation of allele frequency spectrum from neutral expectation | Genome-wide or per-window values; negative values suggest selection or expansion |
| iHS or XP-EHH | Haplotype-based selection scans | Manhattan plot with significance threshold |
| ROH distribution | Runs of homozygosity — inbreeding, population bottlenecks | Per-sample ROH count and total length distribution |
Figure 3: Diversity and selection metrics dashboard — the six core metrics that transform a population genomics report from descriptive to analytical, providing evidence for evolutionary inference.
Each metric should be accompanied by a brief interpretation note: what the value means in biological terms, how it compares to published values for similar species or populations, and whether any values fall outside expected ranges. Metrics without interpretation are numbers; metrics with interpretation are evidence.
Methods Summary and Interpretation
The final layer of the report is the one you will use most directly: a publication-ready methods section and plain-language interpretation notes.
The methods summary should be written as a draft for your manuscript. It must include:
- Sequencing platform, read configuration, and library preparation kit
- Read trimming and QC tools with version numbers and non-default parameters
- Alignment tool and reference genome assembly accession
- Variant calling pipeline — caller, version, filtering thresholds (QD, MQ, FS, SOR, MQRankSum, ReadPosRankSum), and VQSR or hard-filtering approach
- LD pruning parameters for population structure analyses
- Software and versions for PCA (PLINK, EIGENSOFT), ADMIXTURE, phylogenetic reconstruction, diversity statistics, and selection scans
- Key parameter choices: MAF threshold, missingness threshold, HWE p-value threshold, K values tested in ADMIXTURE
The interpretation notes should explain, in plain language:
- Whether the data quality supports the planned downstream analyses
- Which population structure findings are robust (replicated across methods) versus suggestive (visible in one method only)
- Whether diversity metrics are consistent with known demographic history of the species or population
- Whether any selection scan peaks map to genes with known functions relevant to the study question
- Which results should be validated with additional data or orthogonal methods before publication
A report that includes these elements enables researchers to move directly from data delivery to manuscript preparation — the standard that distinguishes a population genomics analysis provider from a sequencing service. For projects where the data has quality issues, our data rescue guide covers how to recover value from problematic datasets.
What a Good Report Enables
A complete population genomics report does more than document results — it makes your project portable. With a well-structured report, you can take the data to a different bioinformatician for reanalysis, merge it with new samples sequenced two years later, or defend it to a reviewer who questions your variant filtering thresholds. A report that is missing layers — no QC tables, no version numbers, no raw data archive — ties your project to a single provider and a single point in time.
The difference between a data dump and a deliverable is not the sequencing quality — the same instrument produced the reads. It is whether the provider treated your project as a transaction or as a research contribution. The checklist above defines the latter. When discussing deliverables with a provider, reference specific sections of this guide. A provider who can produce them without hesitation has done this before. A provider who says "we usually just send the VCF" has not.
Frequently Asked Questions
Raw data delivery means you receive FASTQ files and are responsible for all downstream processing, analysis, and interpretation. A full bioinformatics report includes alignment, variant calling, QC metrics, population structure analysis, diversity statistics, selection scans, figures, and interpretation — everything you need to move directly to manuscript writing. Raw data costs less upfront but requires in-house bioinformatics expertise. A full report costs more but delivers publication-ready results. The right choice depends on whether you have a population genomics bioinformatician on your team.
Length varies by project scope, but a typical report for a 100–300 sample population genomics project with standard analyses (PCA, ADMIXTURE, FST, diversity, selection scans) runs 15–30 pages of figures, tables, and interpretation text. The report should be long enough to include every metric named in this guide with its interpretation, and no longer. A 5-page report for a 300-sample project is missing content. A 100-page report is likely padded with redundant figures.
Request publication-quality vector formats (PDF, SVG, or EPS) for all main figures, and PNG at 300 DPI for supplementary figures. Vector formats allow journal art departments to adjust fonts, colors, and line weights without degrading image quality. If the provider delivers only PNG screenshots, you will need to regenerate figures before submission — which defeats the purpose of a provider-generated report.
Yes, and you should specify custom analyses during the project scoping phase — not after the report is delivered. Common custom requests include: gene ontology enrichment for selection scan peaks, integration with publicly available datasets (1000 Genomes, gnomAD, species-specific databases), landscape genomics (genotype-environment association), demographic modeling (δaδi, fastsimcoal2), and haplotype network construction. Custom analyses add time and cost but can substantially increase the publication impact of the project.
First, determine whether the quality problems are fatal or manageable. Low coverage in a subset of samples — those samples can be excluded from certain analyses. Batch effects detected and quantified — include batch as a covariate. Systematic missingness at specific genomic regions — these regions may be repetitive or paralogous; exclusion is standard. A report that transparently documents quality problems and recommends specific mitigation strategies is a sign of analytical rigor, not failure. If the provider cannot explain what caused a quality problem or how to address it, that is a different concern — and one that our data rescue guide addresses in detail.
References:
- Alexander DH, Novembre J, Lange K. Fast model-based estimation of ancestry in unrelated individuals. Genome Research. 2009;19(9):1655-1664. doi:10.1101/gr.094052.109
- Weir BS, Cockerham CC. Estimating F-statistics for the analysis of population structure. Evolution. 1984;38(6):1358-1370. doi:10.1111/j.1558-5646.1984.tb05657.x
- Purcell S, Neale B, Todd-Brown K, et al. PLINK: a tool set for whole-genome association and population-based linkage analyses. American Journal of Human Genetics. 2007;81(3):559-575. doi:10.1086/519795
- Korneliussen TS, Albrechtsen A, Nielsen R. ANGSD: analysis of next generation sequencing data. BMC Bioinformatics. 2014;15:356. doi:10.1186/s12859-014-0356-4
- Meisner J, Albrechtsen A. Inferring population structure and admixture proportions in low-depth NGS data. Genetics. 2018;210(2):719-731. doi:10.1534/genetics.118.301336
For Research Use Only. Not for use in diagnostic procedures or clinical decision-making.
Related
Services
Download
