Statistical Genomics Foundry
Paper
domain/whole-genome-duplicationdomain/gene-family-evolution

Qiao et al. 2019 — DupGen_finder (modes of gene duplication in plants)

1. Citation

Qiao X, Li Q, Yin H, Qi K, Li L, Wang R, Zhang S, Paterson AH. “Gene duplication and evolution in recurring polyploidization–diploidization cycles in plants.” Genome Biology 20:38 (2019). DOI: 10.1186/s13059-019-1650-2. PMID 30791939, PMCID PMC6383267. Article license: Creative Commons Attribution 4.0 International (CC-BY 4.0) → verbatim reproduction permitted.

Companion software: DupGen_finder, GitHub qiao-xin/DupGen_finder (https://github.com/qiao-xin/DupGen_finder), a Perl pipeline. README accessed from the master branch on 2026-07-05. Repository code license reported as GPL-3.0. The README does not print an explicit version string; a specific commit hash was not captured [access flag: version/commit not pinned].

2. Access note

Full text read via the open-access PMC copy (PMC6383267); the publisher (Springer/BMC) canonical URL redirected to an auth gate, so PMC was the read surface. CLI defaults and output file naming taken from the raw master README. Paper prose in §7 was extracted through a fetch/summarization tool; the CLI/usage strings in §5–§7 are copied from the raw README file and are trustworthy verbatim functional strings. Paper-prose quotes are close-verbatim but should be re-checked against the PDF before being treated as character-exact.

3. Thesis (1 sentence)

Across 141 sequenced plant genomes, genes arising from five distinct modes of duplication — whole-genome, tandem, proximal, transposed, dispersed — differ systematically in abundance, selection pressure, expression divergence, and gene-conversion rate, and DupGen_finder is the pipeline built to classify genes into these five modes.

4. Problem & context

Plant genomes are shaped by recurring polyploidization–diploidization cycles; duplicate genes originate by several mechanisms with different downstream fates. The paper needed a uniform, automated way to assign every duplicated gene in a genome to exactly one origin mode so that the evolutionary consequences of each mode could be compared across many genomes.

5. Method / approach (what the pipeline actually does)

Five duplication-mode classes (this is the exact set; there is no “segmental” class): WGD, tandem, proximal, transposed, dispersed. Per-species output files are named <sp>.wgd.pairs, <sp>.tandem.pairs, <sp>.proximal.pairs, <sp>.transposed.pairs, <sp>.dispersed.pairs (verbatim functional strings from README).

Classification pipeline / priority order (a gene pair is assigned to the first class it matches; later classes are computed on the remainder):

  1. all-vs-all local BLASTP within the target genome → candidate homologous pairs.
  2. run MCScanX intra-genome → collinear (anchored) pairs = WGD-derived pairs.
  3. remove WGD pairs; among the remaining homologous pairs on the same chromosome:
    • tandem = the two BLAST-hit genes are directly adjacent on the same chromosome.
    • proximal = non-tandem same-chromosome pair separated by ≤ -d genes (default 10).
  4. remove WGD + tandem + proximal; identify transposed pairs using ancestral-vs-non- ancestral locus context, requiring an outgroup: one copy sits in an ancestral locus (parent copy), the other in a non-ancestral locus (transposed copy). Inter-species collinearity is computed by running MCScanX on target↔outgroup BLASTP to establish which loci are ancestral.
  5. everything left after removing WGD + tandem + proximal + transposed = dispersed.

Exact CLI usage and defaults (verbatim from README master):

Usage: DupGen_finder.pl -i data_directory -t target_species -c outgroup_species -o output_directory
#####################
Optional:
-a 1 or 0(are segmental duplicates ancestral loci or not? default: 1, yes)
-d number_of_genes(maximum distance to call proximal, default: 10)
#####################
The following are optional MCScanX parameters:
-k match_score(cutoff score of collinear blocks for MCScanX, default: 50)
-g gap_penalty(gap penalty for MCScanX, default: -1)
-s match_size(number of genes required to call a collinear block for MCScanX, default: 5)
-e e_value(alignment significance for MCScanX, default: 1e-05)
-m max_gaps(maximum gaps allowed for MCScanX, default: 25)
-w overlap_window(maximum distance in terms of gene number, to collapse BLAST matches for MCScanX, default: 5)

Required inputs (tab-delimited; all four in one directory):

  • <target>.gff — gene positions; columns: sp_chr <tab> gene_id <tab> start <tab> end.
  • <target>.blast — self-genome BLASTP, -outfmt 6 (m8) tabular.
  • <target>_<outgroup>.gff — combined GFF for target + outgroup.
  • <target>_<outgroup>.blast — cross-genome (target vs outgroup) BLASTP, -outfmt 6. Paper: all-vs-all BLASTP run with E < 1e-10, top 5 matches, m8 output; for each gene only the transcript with the longest CDS was kept.

6. Key claims / findings (atomic)

  • DupGen_finder classifies every duplicated gene into exactly one of five modes: WGD, tandem, proximal, transposed, dispersed. “segmental” is NOT one of the five output classes.
  • tandem = the two homologous genes are directly adjacent on the same chromosome (0 intervening genes); there is no separate tunable “tandem count” parameter — adjacency is the definition.
  • proximal = non-tandem, same-chromosome pair separated by ≤ 10 genes by default; the cutoff is the -d option, default: 10. There is a single distance parameter, not a tandem/proximal pair of windows.
  • WGD pairs are exactly the intra-genome MCScanX collinear (anchored/syntenic) pairs; small-scale modes are computed only after WGD pairs are removed from the homolog set.
  • transposed classification needs an outgroup to define ancestral vs non-ancestral loci (via inter-species MCScanX synteny); one copy in ancestral locus (parent), one in a non-ancestral locus (transposed copy).
  • dispersed = the residual class: homologous pairs neither collinear, adjacent, proximal, nor transposed.
  • The -a option (default: 1) asks “are segmental duplicates ancestral loci or not?” — this is the only place the word “segmental” appears, and it is a transposed-duplication ancestry assumption toggle, not a class label.

7. Load-bearing statements (CC-BY 4.0 → verbatim allowed)

Functional strings (README master, verbatim, high confidence):

  • -d number_of_genes(maximum distance to call proximal, default: 10)
  • -a 1 or 0(are segmental duplicates ancestral loci or not? default: 1, yes)
  • -w overlap_window(maximum distance in terms of gene number, to collapse BLAST matches for MCScanX, default: 5)
  • Output classes / file suffixes: .wgd.pairs, .tandem.pairs, .proximal.pairs, .transposed.pairs, .dispersed.pairs.

Paper prose (Methods / Abstract, close-verbatim — extracted via fetch tool, re-verify against PDF for character-exactness):

  • Abstract, category list: “Genes derived from whole-genome, tandem, proximal, transposed, or dispersed duplication differ in abundance, selection pressure, expression divergence, and gene conversion rate among genomes.”
  • Tandem definition: “If the two genes in a BLASTP hit that are adjacent to each other on the same chromosome, they were defined as tandem gene pair.”
  • Proximal definition: “Proximal gene pairs were defined as non-tandem pairs separated by 10 or fewer genes on the same chromosome.”
  • Transposed definition: “one gene existed in its ancestral locus (named the parent copy) and the other was located in a non-ancestral locus (transposed copy).”
  • Dispersed definition: “After removing WGD, tandem, proximal, and transposed duplications from the whole set of homologous gene pairs, the remaining gene pairs were classified as dispersed duplications.”

8. Stated scope, assumptions, limitations

  • Designed for and validated on plant genomes (141 sequenced plants).
  • Requires a suitable outgroup genome to resolve transposed duplications and ancestral loci; transposed/dispersed separation depends on that outgroup choice.
  • Depends on MCScanX for both intra- and inter-species collinearity; WGD calls inherit MCScanX’s collinear-block parameters (-k, -g, -s, -e, -m, -w defaults above).
  • Uses only the longest-CDS transcript per gene; alternative isoforms are dropped upstream.

9. Failure modes / invalidity patterns

  • Wrong or missing outgroup → transposed duplications cannot be correctly identified, so genes that are truly transposed fall through into the dispersed residual (over-counting dispersed).
  • -d mis-set shifts the tandem/proximal boundary: raising it reclassifies dispersed/ distant pairs as proximal; there is no separate tandem window, so tandem is fixed at adjacency.
  • MCScanX collinearity params too strict/loose directly change what counts as WGD (the anchored set), which cascades: everything not called WGD is pushed into the small-scale modes.
  • Priority-order dependence: because classes are assigned by subtraction in a fixed order (WGD → tandem → proximal → transposed → dispersed), a pair is counted once in its highest- priority matching class; a gene can appear in multiple pairs of different modes.
  • Symptom of the common downstream error: a skill claiming a class called “segmental” or claiming “tandem = 5 genes, proximal = 5–25 genes” is inconsistent with this tool — the class is “transposed” not “segmental,” tandem = adjacency, and proximal is a single -d-genes window (default 10).

10. What the source does NOT address (confident silences)

  • README gives no explicit prose algorithm for how tandem/proximal/transposed/dispersed are computationally separated (that detail is in the paper Methods, not the README).
  • README carries no explicit version string or changelog captured here.
  • Behavior for non-plant genomes is not characterized.

11. Open questions / ambiguities

  • Whether “adjacent” for tandem strictly means 0 intervening genes vs a small settable gap — the definition read implies directly adjacent (0), and no CLI knob exposes a tandem window.
  • Exact tie-breaking when a gene participates in pairs of several modes (the pipeline is pair- based; per-gene mode assignment across conflicting pairs is not spelled out in what was read).
  • The precise role/effect of -a (segmental duplicates ancestral or not, default 1) on transposed-duplication ancestry inference is stated only tersely in the README.

12. Guidance answers

  1. Exact class set / “transposed” vs “segmental”: The five classes are WGD, tandem, proximal, transposed, dispersed. The fifth is transposed (confirmed). “segmental” is NOT a class label — it appears only in the -a option text (“are segmental duplicates ancestral loci or not? default: 1, yes”). The downstream “segmental” claim is wrong.
  2. Default window/distance: -d number_of_genes(maximum distance to call proximal, default: 10). A single -d default of 10. Tandem is defined by adjacency, not a count parameter; there is no separate tandem-count or separate proximal 5–25 window. The downstream “tandem=5, proximal 5–25” claim is wrong. (Note: -w overlap_window ... default: 5 is an MCScanX BLAST-collapse window, unrelated to tandem/proximal calling.)
  3. WGD vs small-scale separation: MCScanX intra-genome collinearity yields the anchored WGD-derived pairs; these are removed first, then tandem/proximal are called from same- chromosome distance, transposed from ancestral-vs-non-ancestral loci resolved by inter-genome MCScanX synteny against the outgroup, and dispersed is the residual. Verbatim: “The MCScanX algorithm was utilized to identify the WGD-derived gene pairs”; “After removing WGD, tandem, proximal, and transposed duplications from the whole set of homologous gene pairs, the remaining gene pairs were classified as dispersed duplications.”
  4. Required inputs/formats: <target>.gff (sp_chr, gene_id, start, end; tab-delimited), <target>.blast (self BLASTP -outfmt 6), <target>_<outgroup>.gff (combined), and <target>_<outgroup>.blast (cross-genome BLASTP -outfmt 6). MCScanX is the collinearity engine. Paper BLASTP settings: E < 1e-10, top 5 matches, m8; longest-CDS transcript per gene.
  5. Citation/license: Qiao X, Li Q, Yin H, Qi K, Li L, Wang R, Zhang S, Paterson AH, Genome Biology 20:38 (2019), DOI 10.1186/s13059-019-1650-2, article CC-BY 4.0. Confirmed. (Tool repo code license reported GPL-3.0.)