From Reference to Pangenome: How Graph-Based Methods Are Redefining Comparative Genomics

By Dr. Zubair Khalid, DVM, MS, PhD · 2026-08-21

Key Takeaways

• Pangenome graphs represent the complete genetic material of a group of related organisms, including core and accessory genes, by storing diverse haplotypes as a graph structure where nodes are sequence segments and edges represent adjacencies. This contrasts with linear references, which represent a single sequence and limit the capture of population-level allelic and structural variation.
• The shift to pangenome graphs is driven by advances in low-cost whole-genome assembly, enabling the collection of haplotype-resolved genomes and facilitating precise analysis of sequence and variation across diverse populations, particularly for structural variants (insertions, deletions, inversions, duplications >50 bp) that are poorly represented in linear references.
• Construction methods like Minigraph-Cactus and PanGenome Graph Builder leverage whole-genome alignments or all-to-all alignments, respectively, to build graphs that can represent variation at multiple scales, from single nucleotide variants to large structural rearrangements, offering a more comprehensive coordinate system than linear references.
• Graph-based methods, such as HISAT2 for alignment and Panaroo for prokaryotic pangenome clustering, enable read mapping and genotyping against a population of genomes rather than a single reference, improving accuracy for genes with high allelic diversity (e.g., immune system genes) and mitigating annotation errors in prokaryotic assemblies.
• Adopting pangenome graph methods requires careful consideration of research questions, input assembly quality, computational resources, and tool compatibility, with validation steps crucial to ensure graph completeness, accuracy, and absence of bias, especially when dealing with complex genomic regions or diverse populations.

Comparative genomics has long depended on a single linear reference genome as the coordinate system for all downstream analyses. That dependency creates a practical problem for researchers who work with diverse populations, structural variants, or species with high genomic diversity. A linear reference represents only a small number of individuals, which limits its usefulness for genotyping and variant discovery across a population. Pangenome graphs address this limitation by storing a representative set of diverse haplotypes and their alignment, usually as a graph structure. This article explains the conceptual shift from linear references to pangenome graphs, describes construction methods, and outlines how graph-based approaches change comparative genomics workflows. The intended reader is a biology student, researcher, laboratory professional, or life-science practitioner who needs to understand when and how to adopt pangenome graph methods in their own analyses.

The Linear Reference Problem

What a Linear Reference Genome Does and Does Not Represent

A linear reference genome is a single contiguous sequence that serves as the coordinate system for mapping reads, calling variants, and annotating genes. Most researchers have used the human reference genome or a model organism reference in this way. The human reference genome represents only a small number of individuals, which limits its usefulness for genotyping because alleles that are common in some populations may be absent from the reference sequence entirely. When a read contains a sequence that is not present in the reference, the aligner must either place it at a nearby location with mismatches or discard it as unmapped. Both outcomes reduce the accuracy of downstream variant calls and expression measurements.

The same limitation applies to non-human genomes. A single reference for a crop species, livestock breed, or microbial species cannot capture the full allelic diversity present across the populations that researchers actually study. Structural variants, which include insertions, deletions, inversions, and duplications larger than 50 base pairs, are particularly poorly represented in linear references because the reference contains only one version of each locus. If a population carries a large insertion that is absent from the reference, reads from that insertion have no correct alignment target.

How Assembly Advances Changed the Landscape

Low-cost whole-genome assembly has enabled the collection of haplotype-resolved pangenomes for numerous organisms. This technological change is encouraging the development of methods that can precisely address the sequence and variation described in large collections of related genomes. Instead of comparing every new sample to a single reference, researchers can now assemble multiple high-quality genomes from a population and compare them to each other. These approaches often use graphical models of the pangenome to support algorithms for sequence alignment, visualization, functional genomics, and association studies.

The practical consequence is that researchers now have a choice. They can continue using a linear reference and accept the biases that come with it, or they can build or use a pangenome graph that represents the diversity of their study population. The choice depends on the research question, the quality of available assemblies, and the computational resources available.

What Is a Pangenome Graph

Core Concepts and Terminology

A pangenome is the complete set of genetic material across a group of related organisms, typically a species or a clade. The pangenome includes both core genes, which are present in all members of the group, and accessory genes, which are present in only some members. In prokaryotes, accessory gene content can vary substantially due to horizontal gene transfer, gene duplication, and gene loss. In eukaryotes, the pangenome concept captures structural variation and presence-absence variation across individuals or breeds.

A pangenome graph represents the pangenome as a graph structure. In this graph, nodes represent sequence segments and edges represent adjacencies between segments. Paths through the graph represent individual haplotypes. A linear reference is a special case of a pangenome graph in which there is only one path. When multiple haplotypes are included, the graph branches at positions where the haplotypes differ and rejoins where they are identical. This structure allows a single coordinate system to relate multiple sequences without forcing every sample to align to one arbitrary reference.

How Graphs Represent Variation at Different Scales

Pangenome graphs can represent variation at different scales simultaneously. Single nucleotide variants appear as small bubbles in the graph where two or more alternative bases occupy the same position. Structural variants appear as larger bubbles or as alternative paths that traverse different sets of nodes. The ability to represent variation at different scales is a direct benefit of constructing the graph from whole-genome alignments instead of from variant calls alone.

Constructing a pangenome graph directly from assemblies, as opposed to variant calls, leverages the graph's ability to represent variation at different scales. Variant callers typically identify small variants and some larger events, but they may miss complex structural rearrangements that are visible in whole-genome alignments. When the graph is built from assemblies, the aligner can identify all forms of variation because the input sequences contain the full genomic context.

Relationship to Core and Accessory Genomes

Pangenome graphs provide a natural framework for distinguishing core and accessory genomic content. In a graph, core regions appear as nodes that are traversed by all haplotype paths. Accessory regions appear as nodes or subgraphs that are traversed by only a subset of paths. This distinction is important for both prokaryotic and eukaryotic studies. In prokaryotes, accessory genes often carry functions related to antibiotic resistance, virulence, or metabolic versatility. In eukaryotes, accessory regions may include breed-specific or population-specific structural variants that influence phenotype.

The graph structure also makes it possible to measure conservation across the pangenome. Regions that are conserved across all haplotypes have high coverage in the graph, while variable regions have lower coverage or multiple alternative paths. This information can guide downstream analyses such as primer design, functional annotation, and association testing.

Pangenome Graph Construction Methods

Minigraph-Cactus Pipeline

The Minigraph-Cactus pangenome pipeline creates pangenomes directly from whole-genome alignments. This method was demonstrated on 90 human haplotypes from the Human Pangenome Reference Consortium and on a Drosophila melanogaster pangenome. The pipeline builds graphs that contain all forms of genetic variation while allowing use of current mapping and genotyping tools. This compatibility with existing tools is a practical advantage because researchers do not need to replace their entire analysis stack to use the pangenome.

The quality and completeness of reference genomes used for analysis within the pangenomes affects the accuracy of the methods. The developers showed that using the CHM13 reference from the Telomere-to-Telomere Consortium improves the accuracy of their methods. This finding has a direct practical implication: the choice of reference genome matters even when building a pangenome graph. A more complete reference provides better anchors for the graph and reduces the number of misassemblies or gaps that propagate into the graph structure.

PanGenome Graph Builder

The PanGenome Graph Builder is a pipeline for constructing pangenome graphs without bias or exclusion. Current approaches to build pangenome graphs either exclude complex sequences or are based upon a single reference. The PanGenome Graph Builder addresses both limitations by using all-to-all alignments to build a variation graph. In this graph, researchers can identify variation, measure conservation, detect recombination events, and infer phylogenetic relationships.

The all-to-all alignment strategy is computationally more demanding than reference-based approaches because every input assembly must be aligned to every other input assembly. However, this approach avoids the bias that comes from choosing one assembly as the reference. When a graph is built from all-to-all alignments, no single haplotype is privileged over the others. This property is valuable for studies where the choice of reference could influence the results, such as population genetics or comparative genomics across divergent lineages.

HISAT2 and HISAT-genotype

HISAT2 is a method that can align both DNA and RNA sequences using a graph Ferragina Manzini index. The method represents and searches an expanded model of the human reference genome in which over 14.5 million genomic variants in combination with haplotypes are incorporated into the data structure used for searching and alignment. This approach allows the aligner to map reads to a population of genomes instead of to a single reference.

HISAT-genotype extends this capability to haplotype-resolved genes or genomic regions. The software was applied to HLA typing and DNA fingerprinting, and it outperforms other computational methods and matches or exceeds the performance of laboratory-based assays. For researchers working on genes with high allelic diversity, such as immune system genes, this approach provides a practical alternative to laboratory-based genotyping.

Panaroo for Prokaryotic Pangenomes

Panaroo is a graph-based pangenome clustering tool that accounts for many of the sources of error introduced during the annotation of prokaryotic genome assemblies. Population-level comparisons of prokaryotic genomes must take into account the substantial differences in gene content resulting from horizontal gene transfer, gene duplication, and gene loss. However, the automated annotation of prokaryotic genomes is imperfect, and errors due to fragmented assemblies, contamination, diverse gene families, and mis-assemblies accumulate over the population. These errors have profound consequences when analysing the set of all genes found in a species.

Panaroo addresses these issues by using a graph structure that can identify and correct annotation errors. The tool is available at https://github.com/gtonkinhill/panaroo. For researchers working with bacterial or archaeal genomes, Panaroo provides a way to build pangenomes that are less sensitive to the quality of individual genome annotations.

At a Glance

Practical Workflow for Pangenome Analysis

Step 1: Define the Research Question and Sample Set

The first decision is whether a pangenome graph is necessary for the research question. If the study involves a single genome or a small number of closely related genomes, a linear reference may be sufficient. If the study involves a population with known structural variation, presence-absence variation, or high allelic diversity, a pangenome graph is likely to provide better results.

The sample set must include enough high-quality assemblies to represent the diversity of the study population. For prokaryotes, this may mean dozens to hundreds of genomes. For eukaryotes, the number is typically smaller due to assembly cost, but the assemblies must be haplotype-resolved to capture allelic diversity. The quality of the input assemblies directly affects the quality of the pangenome graph.

Step 2: Assess Input Assembly Quality

Before building a pangenome graph, assess the quality of each input assembly. Key metrics include contiguity, completeness, and contamination. Fragmented assemblies and mis-assemblies introduce errors into the graph that can propagate through downstream analyses. For prokaryotic pangenomes, annotation errors due to fragmented assemblies and contamination are a known source of error that accumulates over the population.

The quality and completeness of reference genomes used for analysis within the pangenomes affects the accuracy of the methods. If the study includes a reference genome, choose the most complete and accurate reference available. The CHM13 reference from the Telomere-to-Telomere Consortium improved the accuracy of Minigraph-Cactus pangenomes compared to older references.

Step 3: Choose a Construction Method

The choice of construction method depends on the research question, the input data, and the computational resources available. Minigraph-Cactus is appropriate for building pangenomes from whole-genome alignments of eukaryotic genomes, particularly when the goal is to represent all forms of genetic variation while maintaining compatibility with existing mapping and genotyping tools. The PanGenome Graph Builder is appropriate when the goal is to avoid reference bias and when all-to-all alignments are computationally feasible. HISAT2 and HISAT-genotype are appropriate for read alignment and genotyping against a graph that incorporates known variants and haplotypes. Panaroo is appropriate for prokaryotic pangenome clustering when annotation errors are a concern.

Step 4: Build the Graph and Validate

Build the graph using the chosen method and validate the output. Validation steps include checking that all input haplotypes are represented as paths in the graph, that the graph contains the expected variants, and that the graph is free of obvious artifacts such as spurious cycles or disconnected components. For graphs built from whole-genome alignments, check that the alignment boundaries are sensible and that no input assembly is overrepresented or underrepresented.

Step 5: Align Reads and Call Variants

Once the graph is built, align reads to the graph and call variants. The choice of aligner depends on the graph format and the construction method. Minigraph-Cactus graphs allow use of current mapping and genotyping tools, which simplifies this step. HISAT2 provides graph-based alignment for both DNA and RNA sequences. The additional information provided to these methods by the pangenome allows them to achieve superior performance on a variety of bioinformatic tasks, including read alignment, variant calling, and genotyping.

Step 6: Interpret Results in the Graph Context

Interpretation of results must account for the graph structure. Variants are called relative to the graph, not relative to a single linear reference. This means that a variant may be present in some haplotypes and absent in others, and the graph provides the context for understanding this distribution. For core and accessory genome analysis, use the graph to identify which regions are present in all haplotypes and which are variable.

Options and Tradeoffs in Graph Construction

Reference-Based versus All-to-All Alignment

Reference-based graph construction aligns all input assemblies to a single reference and incorporates variants into the graph. This approach is computationally efficient and works well when a high-quality reference is available. However, it inherits the biases of the reference. If the reference is missing sequences that are common in the study population, those sequences will be absent from the graph.

All-to-all alignment avoids reference bias by aligning every input assembly to every other input assembly. The PanGenome Graph Builder uses this strategy to build a variation graph without bias or exclusion. The tradeoff is computational cost. All-to-all alignment scales quadratically with the number of input assemblies, so it becomes expensive for large sample sets.

Variant Calls versus Assembly Input

Pangenome graphs can be constructed from variant calls or from assemblies. Alternate alleles determined by variant callers can be used to construct pangenome graphs, but advances in long-read sequencing are leading to widely available, high-quality phased assemblies. Constructing a pangenome graph directly from assemblies, as opposed to variant calls, leverages the graph's ability to represent variation at different scales.

Variant calls are limited by what the variant caller can detect. Small variants are usually well represented, but complex structural variants may be missed. Assemblies contain the full genomic sequence, so graphs built from assemblies can represent all forms of variation. The tradeoff is that assemblies are more expensive to produce and require more computational resources to align.

Graph Size and Complexity

Pangenome graphs can become large and complex, especially when built from many high-quality assemblies. The graph size affects memory usage, alignment speed, and downstream analysis performance. Researchers should consider whether their computational infrastructure can support the graph size before committing to a construction method.

The Minigraph-Cactus pipeline was demonstrated on 90 human haplotypes, which is a substantial graph. The developers showed that the method scales to this size while allowing use of current mapping and genotyping tools. For larger sample sets, researchers may need to subsample haplotypes or use a more memory-efficient graph representation.

Observations and Measurements in Pangenome Analysis

Measuring Graph Quality

Graph quality can be measured by several metrics. Completeness measures whether all input haplotypes are represented as paths in the graph. Accuracy measures whether the graph correctly represents the known variation in the input assemblies. Bias measures whether any single haplotype or reference dominates the graph structure.

The quality and completeness of reference genomes used for analysis within the pangenomes affects the accuracy of the methods. This observation from the Minigraph-Cactus study has a practical implication: researchers should document which reference was used and how it affected the graph. If a more complete reference becomes available, rebuilding the graph may improve accuracy.

Measuring Conservation and Variation

Pangenome graphs provide a natural framework for measuring conservation and variation. Node coverage across haplotypes indicates how conserved a region is. Regions traversed by all haplotypes are core regions. Regions traversed by only some haplotypes are accessory or variable regions. The PanGenome Graph Builder allows researchers to identify variation, measure conservation, detect recombination events, and infer phylogenetic relationships from the graph.

These measurements can be reported as summary statistics for the pangenome. For example, the number of core nodes, the number of accessory nodes, and the distribution of node coverage across haplotypes provide a quantitative description of the pangenome structure.

Recording Analysis Parameters

Reproducibility requires recording all analysis parameters. For pangenome construction, record the construction method, the version of the software, the input assemblies and their quality metrics, the reference genome if one was used, and all alignment parameters. For downstream analyses, record the aligner, the variant caller, and all parameters used.

The Galaxy Training Network provides accessible workflow training and analysis tutorials that emphasize reproducibility. The nf-core documentation describes community pipeline standards, usage, configuration, and reproducible workflow context. These resources can help researchers structure their pangenome analyses for reproducibility.

Records and Documentation

What to Record for Each Pangenome Analysis

Maintain a record of the following items for each pangenome analysis:

1. Input assemblies, including source, version, and quality metrics
2. Construction method and software version
3. Reference genome used, if any, and its version
4. Alignment parameters and any filtering thresholds
5. Graph statistics, including number of nodes, edges, and paths
6. Validation results, including completeness and accuracy checks
7. Downstream analysis parameters and software versions
8. Date of analysis and person responsible

This record supports reproducibility and allows other researchers to understand the decisions that shaped the pangenome graph.

Documentation Standards

The Carpentries Lessons provide foundational training in computing, data, shell, Git, and programming that supports good documentation practices. Version control for analysis scripts and parameters is essential for reproducible pangenome analysis. The nf-core documentation describes community pipeline standards that include configuration and usage documentation. Adopting these standards makes it easier to share pangenome analyses with collaborators and to publish them with confidence.

Sharing and Publication

When publishing pangenome analyses, share the graph itself, beyond the summary statistics. Graphs can be large, so consider using a public repository or database. The NCBI Data Resources provide official descriptions of NCBI databases, search systems, sequence resources, and analysis services that may be appropriate for sharing input assemblies and some graph outputs. The EMBL-EBI Training provides bioinformatics learning pathways and data-resource training that can help researchers understand how to deposit and access genomic data.

Quality Controls and Validation

Input Quality Control

Quality control begins with the input assemblies. Check each assembly for completeness, contiguity, and contamination before including it in the pangenome. Fragmented assemblies and mis-assemblies introduce errors into the graph. For prokaryotic genomes, contamination is a particular concern because it can introduce genes from other species into the pangenome.

The Panaroo pipeline was designed to account for many of the sources of error introduced during the annotation of prokaryotic genome assemblies. Errors due to fragmented assemblies, contamination, diverse gene families, and mis-assemblies accumulate over the population, leading to profound consequences when analysing the set of all genes found in a species. Using a tool like Panaroo can mitigate these errors.

Graph Validation

After building the graph, validate that it correctly represents the input assemblies. Check that each input haplotype is present as a path in the graph. Check that known variants are represented. Check for spurious structures such as cycles that do not correspond to biological variation.

For graphs built from whole-genome alignments, check the alignment quality. Misalignments can create false variants or obscure real ones. The Minigraph-Cactus pipeline was designed to build graphs containing all forms of genetic variation while allowing use of current mapping and genotyping tools. Validation should confirm that these properties hold for the specific graph.

Downstream Validation

Validate downstream analyses by comparing results to known variants or to results from a linear reference analysis. If the pangenome graph is working correctly, it should identify variants that are missed by linear reference analysis, particularly structural variants and presence-absence variants. The additional information provided to these methods by the pangenome allows them to achieve superior performance on a variety of bioinformatic tasks, including read alignment, variant calling, and genotyping.

Common Failure Patterns

Reference Bias Persists in Graph Construction

A common failure is assuming that building a pangenome graph eliminates all reference bias. If the graph is built using a reference-based approach, the reference still influences the graph structure. Sequences that are absent from the reference may be absent from the graph. The PanGenome Graph Builder was developed specifically to address this problem by using all-to-all alignments to build a variation graph without bias or exclusion.

Poor Input Assemblies Produce Poor Graphs

Another common failure is using low-quality input assemblies. Fragmented assemblies, mis-assemblies, and contamination all introduce errors into the graph. These errors accumulate over the population and can lead to incorrect conclusions about core and accessory genome content. The Panaroo pipeline was designed to account for these errors in prokaryotic pangenomes, but the best approach is to use high-quality assemblies from the start.

Incompatible Tools and Formats

A third failure pattern is using tools that are not compatible with the graph format. Not all aligners and variant callers can work with pangenome graphs. The Minigraph-Cactus pipeline was designed to allow use of current mapping and genotyping tools, but other graph formats may require specialized tools. Check tool compatibility before committing to a construction method.

Overinterpreting Graph Structure

A fourth failure pattern is overinterpreting the graph structure. The graph represents the input assemblies, and its properties are limited by the quality and diversity of those assemblies. A graph built from a small number of closely related genomes will not represent the full diversity of a species. Researchers should be careful not to generalize from a limited pangenome to the entire species.

Limitations of Pangenome Graphs

Computational Cost

Pangenome graphs are computationally expensive to build and analyze. All-to-all alignment scales quadratically with the number of input assemblies. Even reference-based approaches require substantial memory and storage for large graphs. Researchers with limited computational resources may need to subsample haplotypes or use simpler approaches.

Dependence on Input Quality

The quality of the pangenome graph depends entirely on the quality of the input assemblies. Errors in the input assemblies propagate into the graph. The quality and completeness of reference genomes used for analysis within the pangenomes affects the accuracy of the methods. Researchers must invest in high-quality assemblies to build high-quality graphs.

Unclear Whether Graphs Will Replace Linear References

It is unclear whether pangenome graphs will replace linear reference genomes. Their ability to harmoniously relate multiple sequence and coordinate systems will make them useful irrespective of which pangenomic models become most common in the future. For now, researchers may need to work with both linear references and pangenome graphs, using each for the tasks where it is most appropriate.

Annotation Errors in Prokaryotic Genomes

For prokaryotic pangenomes, automated annotation is imperfect. Errors due to fragmented assemblies, contamination, diverse gene families, and mis-assemblies accumulate over the population. These errors have profound consequences when analysing the set of all genes found in a species. Tools like Panaroo can mitigate these errors, but they cannot eliminate them entirely.

Safety and Regulatory Context

Data Management and Privacy

Pangenome analyses often involve genomic data from human subjects or from agricultural species. For human data, researchers must comply with privacy regulations and data-sharing policies. The NCBI Data Resources provide official descriptions of databases and search systems that may have specific data-sharing requirements. Researchers should review these requirements before depositing or accessing data.

Reproducibility Requirements

Many journals and funding agencies require reproducible analyses. The Galaxy Training Network provides accessible workflow training and analysis tutorials that emphasize reproducibility. The nf-core documentation describes community pipeline standards that support reproducible workflow context. Adopting these standards can help researchers meet reproducibility requirements.

Professional Escalation Criteria

Researchers should escalate to a more experienced colleague or a bioinformatics core facility when they encounter any of the following situations:

1. The pangenome graph is too large for available computational resources
2. Input assemblies have quality issues that cannot be resolved with available tools
3. The graph construction method produces unexpected or inconsistent results
4. Downstream analyses produce results that conflict with known biology
5. The research question requires expertise beyond the current team's capabilities

Decision Framework for Choosing Between Linear References and Pangenome Graphs

Structured Assessment Before Committing Resources

Researchers often commit to a pangenome graph project without a structured evaluation of whether the investment will improve their specific analyses. A practical decision framework helps avoid wasted compute time and storage. The framework below uses four assessment domains that map directly to the documented strengths and limitations of pangenome graphs.

Domain 1: Population Diversity Scope

Evaluate whether the study population carries variation that a single linear reference cannot represent. The human reference genome represents only a small number of individuals, which limits its usefulness for genotyping. The same limitation applies to any species where a single reference cannot capture population diversity. Ask whether the study involves structural variants larger than 50 base pairs, presence-absence variation, or alleles that are common in the study population but absent from the reference. If the answer is yes to any of these, a pangenome graph is likely to provide better results than a linear reference.

For prokaryotic studies, assess gene content variability. Population-level comparisons of prokaryotic genomes must take into account the substantial differences in gene content resulting from horizontal gene transfer, gene duplication, and gene loss. If the species is known to have high accessory gene content, a pangenome approach is indicated.

Domain 2: Input Data Readiness

Pangenome graphs require high-quality assemblies from multiple individuals. Advances in long-read sequencing are leading to widely available, high-quality phased assemblies. Assess whether the available assemblies are haplotype-resolved and complete. Fragmented assemblies and mis-assemblies introduce errors into the graph that accumulate over the population. The quality and completeness of reference genomes used for analysis within the pangenomes affects the accuracy of the methods.

For prokaryotic genomes, automated annotation is imperfect, and errors due to fragmented assemblies, contamination, diverse gene families, and mis-assemblies accumulate over the population. If the input assemblies have known quality issues, address those before building a graph. The Panaroo pipeline can account for many annotation errors, but it cannot compensate for poor assembly quality.

Domain 3: Computational Infrastructure

Pangenome graph construction is computationally demanding. All-to-all alignment scales quadratically with the number of input assemblies. Even reference-based approaches require substantial memory and storage for large graphs. The Minigraph-Cactus pipeline was demonstrated on 90 human haplotypes, which is a substantial graph. Assess whether the available computational infrastructure can support the expected graph size before committing to a construction method.

Consider the downstream analysis requirements as well. The additional information provided to pangenome methods allows them to achieve superior performance on a variety of bioinformatic tasks, including read alignment, variant calling, and genotyping. However, these methods require compatible tools and sufficient memory for alignment and variant calling against the graph.

Domain 4: Analysis Compatibility

Check whether the downstream analysis tools are compatible with the graph format. The Minigraph-Cactus pipeline builds graphs containing all forms of genetic variation while allowing use of current mapping and genotyping tools. HISAT2 can align both DNA and RNA sequences using a graph Ferragina Manzini index. Other graph formats may require specialized tools that are not part of the standard analysis stack.

Scoring Matrix for the Decision Framework

Use the following scoring system to evaluate each domain. Score each domain from 1 to 5, where 1 means the condition strongly favors a linear reference and 5 means the condition strongly favors a pangenome graph.

A total score of 16 or higher indicates that a pangenome graph is likely to provide meaningful benefits. A score of 8 or lower indicates that a linear reference is probably sufficient. Scores between 9 and 15 require careful consideration of the specific research question and the resources available.

Record System for Pangenome Project Decisions

Maintain a decision record that documents the assessment for each domain. This record supports reproducibility and helps other researchers understand why a pangenome approach was or was not chosen. The record should include the following fields:

1. Research question and expected variant types
2. Number of input assemblies and their quality metrics
3. Construction method considered and its requirements
4. Computational resources available and estimated requirements
5. Tool compatibility assessment for downstream analyses
6. Decision outcome with rationale
7. Date of assessment and person responsible

The Galaxy Training Network provides accessible workflow training and analysis tutorials that emphasize reproducibility. The nf-core documentation describes community pipeline standards, usage, configuration, and reproducible workflow context. These resources can help structure the decision record and the subsequent analysis.

Troubleshooting Method for Graph Construction Failures

When graph construction fails or produces unexpected results, use a systematic troubleshooting approach instead of restarting with different parameters blindly.

Step 1: Isolate the Input Problem

Check whether the failure is caused by a specific input assembly. Remove assemblies one at a time and rebuild the graph to identify problematic inputs. For prokaryotic genomes, check for contamination, which can introduce genes from other species into the pangenome. The Panaroo pipeline was designed to account for many of the sources of error introduced during the annotation of prokaryotic genome assemblies.

Step 2: Verify Alignment Quality

For graphs built from whole-genome alignments, check the alignment quality. Misalignments can create false variants or obscure real ones. The PanGenome Graph Builder uses all-to-all alignments to build a variation graph in which researchers can identify variation, measure conservation, detect recombination events and infer phylogenetic relationships. If the graph contains spurious cycles or disconnected components, the alignments may be the source of the problem.

Step 3: Assess Reference Impact

If the construction method uses a reference genome, test whether the reference choice affects the failure. The quality and completeness of reference genomes used for analysis within the pangenomes affects the accuracy of the methods. Using the CHM13 reference from the Telomere-to-Telomere Consortium improved the accuracy of Minigraph-Cactus pangenomes compared to older references. If the graph quality improves with a different reference, the original reference was likely the limiting factor.

Step 4: Document the Resolution

Record the troubleshooting steps and the resolution in the project record. This documentation helps future projects avoid the same issues and provides context for interpreting the final graph.

Common Failure Patterns in the Decision Process

Pattern 1: Overcommitting to a Graph Without Assessing Input Quality

Researchers sometimes build a pangenome graph from low-quality assemblies because they assume the graph will compensate for input errors. This assumption is incorrect. Errors in the input assemblies propagate into the graph and accumulate over the population. The best approach is to use high-quality assemblies from the start.

Pattern 2: Choosing a Construction Method Based on Familiarity instead of Fit

The choice of construction method should follow the research question and input data. Minigraph-Cactus is appropriate for building pangenomes from whole-genome alignments of eukaryotic genomes. The PanGenome Graph Builder is appropriate when the goal is to avoid reference bias. HISAT2 and HISAT-genotype are appropriate for read alignment and genotyping against a graph that incorporates known variants. Panaroo is appropriate for prokaryotic pangenome clustering when annotation errors are a concern. Choosing a method because it is familiar instead of because it fits the data can lead to suboptimal results.

Pattern 3: Ignoring Tool Compatibility Until After Graph Construction

Tool compatibility should be assessed before building the graph, not after. The Minigraph-Cactus pipeline builds graphs that allow use of current mapping and genotyping tools. Other graph formats may require specialized tools. Check tool compatibility before committing to a construction method to avoid building a graph that cannot be analyzed with the available tools.

Pattern 4: Failing to Document Decisions

Without a decision record, other researchers cannot understand why a pangenome approach was chosen or how the graph was constructed. This lack of documentation undermines reproducibility and makes it difficult to interpret the results. The Carpentries Lessons provide foundational training in computing, data, shell, Git, and programming that supports good documentation practices.

Professional Escalation Criteria for the Decision Process

Escalate to a more experienced colleague or a bioinformatics core facility when any of the following situations arise during the decision process:

1. The scoring matrix produces a borderline score between 9 and 15 and the research team cannot agree on the appropriate approach
2. Input assemblies have quality issues that cannot be resolved with available tools
3. The computational requirements exceed available resources and the team is uncertain how to reduce them
4. Tool compatibility assessment reveals that no available aligner or variant caller works with the preferred graph format
5. The research question requires expertise beyond the current team's capabilities in pangenome analysis

The EMBL-EBI Training provides bioinformatics learning pathways and data-resource training that can help researchers build the skills needed for pangenome analysis. The Bioconductor project provides official package, workflow, installation, and reproducible genomic-analysis documentation that may be useful for downstream analysis of pangenome graphs.

Frequently Asked Questions

What is the difference between a linear reference genome and a pangenome graph?

A linear reference genome is a single contiguous sequence that serves as the coordinate system for mapping reads and calling variants. A pangenome graph stores a representative set of diverse haplotypes and their alignment, usually as a graph structure. In the graph, nodes represent sequence segments and edges represent adjacencies between segments. Paths through the graph represent individual haplotypes. The graph can represent variation at different scales, from single nucleotide variants to large structural variants.

When should I use a pangenome graph instead of a linear reference?

Use a pangenome graph when the study population has substantial genetic diversity that is not captured by a single reference genome. This includes studies of structural variation, presence-absence variation, and populations with high allelic diversity. Pangenome graphs are also useful for prokaryotic studies where gene content varies substantially due to horizontal gene transfer. For single-genome studies or studies of closely related genomes, a linear reference may be sufficient.

What input data do I need to build a pangenome graph?

You need high-quality genome assemblies from multiple individuals or strains representing the diversity of the study population. For eukaryotes, haplotype-resolved assemblies are preferred. For prokaryotes, assemblies should be checked for contamination and completeness. The quality and completeness of the input assemblies directly affects the quality of the pangenome graph. Some methods can also build graphs from variant calls, but graphs built directly from assemblies can represent variation at different scales.

Which pangenome graph construction method should I choose?

The choice depends on your research question and computational resources. Minigraph-Cactus builds pangenomes directly from whole-genome alignments and scales to large sample sets while allowing use of current mapping and genotyping tools. The PanGenome Graph Builder uses all-to-all alignments to avoid reference bias. HISAT2 and HISAT-genotype are appropriate for read alignment and genotyping against a graph that incorporates known variants. Panaroo is appropriate for prokaryotic pangenome clustering when annotation errors are a concern.

How does the choice of reference genome affect pangenome graph accuracy?

The quality and completeness of reference genomes used for analysis within the pangenomes affects the accuracy of the methods. Using a more complete reference, such as the CHM13 reference from the Telomere-to-Telomere Consortium, improves the accuracy of pangenome construction compared to older references. Even when building a pangenome graph, the choice of reference matters because it provides anchors for the graph and affects which sequences are included.

Can I use my existing alignment and variant calling tools with pangenome graphs?

Compatibility depends on the graph format and construction method. The Minigraph-Cactus pipeline builds graphs that allow use of current mapping and genotyping tools. HISAT2 uses a graph Ferragina Manzini index to align both DNA and RNA sequences. Other graph formats may require specialized tools. Check tool compatibility before committing to a construction method.

How do pangenome graphs help with core and accessory genome analysis?

Pangenome graphs provide a natural framework for distinguishing core and accessory genomic content. Core regions appear as nodes that are traversed by all haplotype paths. Accessory regions appear as nodes or subgraphs that are traversed by only a subset of paths. This distinction is important for prokaryotic studies where accessory genes often carry functions related to antibiotic resistance or virulence, and for eukaryotic studies where accessory regions may include population-specific structural variants.

What are the main limitations of pangenome graphs?

Pangenome graphs are computationally expensive to build and analyze, especially when using all-to-all alignment. The quality of the graph depends entirely on the quality of the input assemblies. It is unclear whether pangenome graphs will replace linear reference genomes, so researchers may need to work with both. For prokaryotic genomes, annotation errors can accumulate over the population and affect pangenome analysis, though tools like Panaroo can mitigate these errors.

Related Bioinformatics Guides

• Pangenome Graph Construction for Bacterial Genomics
• Single-Cell Sequencing Methods: A Comparative Overview
• Genomic Data Analysis Tools: A Comparative Guide for Researchers
• Benchmarking Atlas-Level Data Integration in Single-Cell Genomics: Methods and Best Practices
• Binning in Metagenomics: From Contigs to Genomes

Related Clinical & Scientific Guides

• A Practical Guide to Detecting Antimicrobial Resistance Genes in Shotgun Metagenomic Data
• Computational Immunology: Modeling the Immune System
• How to Set Hard Filters for Germline Variant Calling: A Practical Guide to GATK Best Practices

References and Further Reading

• NCBI Data Resources. National Center for Biotechnology Information.
• EMBL-EBI Training. European Bioinformatics Institute.
• Bioconductor. Bioconductor Project.
• Galaxy Training Network. Galaxy Project.
• nf-core Documentation. nf-core.
• The Carpentries Lessons. The Carpentries.
• Pangenome graph construction from genome alignments with Minigraph-Cactus.. Nature biotechnology, 2024.
• Building pangenome graphs.. Nature methods, 2024.
• Graph-based genome alignment and genotyping with HISAT2 and HISAT-genotype.. Nature biotechnology, 2019.
• Pangenome Graphs.. Annual review of genomics and human genetics, 2020.
• Producing polished prokaryotic pangenomes with the Panaroo pipeline.. Genome biology, 2020.

This article is educational and does not replace validated analysis plans, institutional policy, clinical interpretation, or specialist review.