BGPT: Paper Review: Automatic genome-wide reconstruction of phylogenetic gene trees

Explore by Goal

Quick Explanation Copied

Paper reviewed: Automatic genome-wide reconstruction of phylogenetic gene trees (SYNERGY) — a bottom-up algorithm that jointly builds orthogroups and their gene trees genome-wide by integrating (i) sequence similarity and (ii) conserved gene order (synteny), then scores confidence via a jackknife scheme.

Long Explanation

SYNERGY (Wapinski et al.) — Visual Paper Review

Task: automatically reconstruct genome-wide orthogroups (orthology/paralogy) and their phylogenetic gene trees, leveraging a known species phylogeny and integrating sequence similarity + synteny + an explicit duplication/loss-informed rooting strategy.

Field	What the paper reports
Core contribution	Scalable bottom-up orthogroup + gene-tree reconstruction with synteny-aware similarity and duplication/loss-informed rooting.
Species tested	9 Ascomycota fungal genomes spanning ~150 million years and including a WGD within the clade.
Gene-family outputs	5282 non-singleton orthogroups; 48,265 genes assigned to non-singleton orthogroups; 4,432 genes to singleton orthogroups.
Quality/confidence evaluation	Jackknife holdouts over species content and gene content; confidence measures assess soundness (no migrating orthology pairs) and completeness (no orthology migration into group).

Genome coverage of orthogroups (9 fungi)

Orthogroup counts: non-singleton vs singleton

Jackknife robustness summary (holdout confidence > 0.9)

Simulated orthology relations (sensitivity vs specificity)

The paper reports simulated performance where SYNERGY achieved 87.1% sensitivity at 98.6% specificity, compared with INPARANOID reporting 69% sensitivity at 99.2% specificity (for pairwise orthologous relations in simulated orthogroups).

Pipeline logic (as described in the paper)

Bottom-up recursion (species tree nodes from leaves to root):

Build/maintain a sparse directed gene similarity graph from all-vs-all FASTA alignments; include edges based on FASTA E-value thresholding and best-hit/identity criteria; weight edges using combined peptide + synteny distance scores.
At each internal species node x, match orthogroups from the two child branches using reciprocal edges + transitive closure to form candidate orthogroups.
Reconstruct gene trees for each candidate orthogroup using Neighbor-Joining on distance matrices; then root using a composite score balancing peptide similarity, synteny similarity, and Poisson duplication/loss likelihood.
If rooting violates orthogroup soundness, iteratively break orthogroups at the root to ensure orthogroups correspond to a single ancestral gene.
Update the gene similarity graph by contracting genes into orthogroups and updating distances using a Neighbor-Joining-like update rule.

Critical appraisal (skeptical, evidence-based)

1) Strengths supported by the paper’s reported evidence

Genome-wide automation with explicit gene-tree output. Unlike many hit-clustering approaches, SYNERGY reconstructs gene trees for each orthogroup and uses those trees to infer duplication/loss history at genomic scale.
Orthogroup definition and enforcement. The paper formalizes orthogroups (soundness/completeness) and includes an explicit "break" mechanism when rooting would imply a duplication earlier than allowed.
Robustness assessment via jackknife. The reported confidence results indicate that most non-singleton orthogroups have very high confidence under gene/species holdouts (79% all-four >0.9; 99% with at least one >0.9).

2) Main methodological sensitivities / potential blind spots

Greedy bottom-up reconstruction. The paper acknowledges the bottom-up procedure can occasionally misassign genes, and suggests that adding top-down reassignment could relax greediness.
Rooting depends on multiple coupled signals and model assumptions. Rooting uses peptide similarity + synteny similarity + a Poisson model for duplication/loss likelihood under inferred duplication/loss counts; if any component is systematically biased, rooting could shift.
Graph edge inclusion can create spurious candidate orthogroups. The paper explicitly justifies permissive edge inclusion (to support completeness) and then relies on later phylogenetic refinement and orthogroup breaking to restore soundness.
Synteny utility may vary by lineage. The paper notes that synteny contributes substantially for this yeast clade but is also applicable when chromosomal structure is not conserved (they reference 'data not shown'). That implies performance may depend on genome architecture and annotation.

3) Evaluation design: what is strong vs what is not fully demonstrated

Strong: the jackknife design evaluates robustness of both soundness and completeness to perturbations in which genes/species are hidden.
Potentially weaker / under-specified: the provided excerpt does not clearly state the full data availability/case-by-case replication details (e.g., whether parameter choices and hyperparameters are fully specified beyond what is in text). The paper itself does not provide a data-availability statement in the included text excerpt, so I cannot verify external reproducibility beyond the methods described here.

What would most disprove SYNERGY’s key claims?

The paper’s central claims are (i) improved orthology/paralogy resolution and gene-tree reconstructions relative to simpler approaches, and (ii) robustness to input perturbations. A skeptical falsification plan should therefore target: (a) orthogroup soundness/completeness under alternative gene prediction sources, (b) rooting stability when synteny signal is weak or absent, and (c) sensitivity to the specific edge inclusion thresholds and composite score weights used in rooting.

Author reviews (click-through)

Feedback:

Updated: March 29, 2026

BGPT Paper Review

Study Novelty

90%

The paper’s novelty is the combination of (i) genome-wide recursive orthogroup construction constrained by a species-tree topology, (ii) simultaneous gene-tree reconstruction per orthogroup, and (iii) a rooting procedure that explicitly integrates peptide similarity, synteny similarity, and a duplication/loss likelihood term while enforcing orthogroup soundness via breaking.

Scientific Quality

90%

High internal coherence: explicit formal definitions, a clear bottom-up recursion, and quantitative robustness assessment via jackknife holdouts. The excerpted text does acknowledge greedy misassignment risk and relies on multiple modeling/signal components (graph edges, rooting Poisson assumptions, synteny contribution), but the reported confidence and benchmark figures support the main claims. Reproducibility is limited by the lack of an explicit data-availability statement in the provided content.

Study Generality

80%

The method is stated to apply to any set of sequenced eukaryotic species with a known phylogeny, but evidence presented is limited to nine fungal Ascomycota genomes where a WGD and synteny signals are central; the paper also notes applicability when chromosomal structure is not conserved, but does so with 'data not shown' in the excerpt.

Study Usefulness

90%

It directly outputs orthogroups and rooted gene trees at genome scale, plus per-orthogroup confidence scores from defined holdout perturbations, enabling systematic comparative genomics and WGD/deduplication analyses.

Study Reproducibility

70%

The provided content specifies many key methodological choices (edge thresholds conceptually, peptide/synteny weighting strategy, and jackknife holdout design), but the excerpt does not include an explicit data-availability statement, so full independent reproduction of results cannot be verified from what is shown.

Explanatory Depth

90%

The paper explains the pipeline end-to-end with explicit mathematical components for rooting (composite score + Poisson likelihood) and for confidence scoring (soundness/completeness under holdouts).

🎁 Authors: Collect 500 Free Science Tokens (≈ $50.0 USD)

Claim My Author Tokens

Use for 125 days of free BGPT access (4 tokens = 1 day) or trade/sell (≈ $50.0 USD)