Bio Comparative Genomics Synteny Analysis

Version Compatibility

Reference examples tested with: MCScanX 1.0+ (wyp1125/MCScanX commit 2020+), JCVI 1.4.21+ (Python port of MCScan), GENESPACE 1.4.0+ (Lovell 2022 eLife 11:e78526), SyRI 1.7.1+ (Goel 2019 Genome Biol 20:277), plotsr 1.1.1+, AnchorWave 1.2.5+ (Song 2022 PNAS 119:e2113075119), i-ADHoRe 3.0.01+, SynNet (Zhao 2017 Plant Cell 29:1278), ntSynt 1.0.4+ (2024), minimap2 2.28+, MUMmer 4.0.0+, OrthoFinder 3.0+, R 4.4+. plotsr requires pysam 0.22+ and seaborn 0.13+.

Before using code patterns, verify installed versions match. If versions differ:

• CLI: MCScanX -h; syri --version; python -m jcvi.compara.catalog ortholog --help
• R: packageVersion('GENESPACE'); ?run_genespace
• Python: pip show jcvi

If code throws MCScanX: argument bad format, syri: input alignment file missing required columns, or GENESPACE: GFF parse error, these tools have brittle input parsing: MCScanX requires 4-column species_chr gene start end BED (non-standard), JCVI expects 4-column simple BED, GENESPACE requires GFF3 with gene feature type. Pre-process with jcvi.formats.gff bed or custom AWK.

Synteny Analysis

"Compare genome architecture between these species" -> Detect conserved gene order (synteny) and infer rearrangement history. Synteny is NOT the same as collinearity: synteny is "genes on same chromosome", collinearity is "same order on same chromosome" (modern usage often conflates them). The choice of tool depends on whether the question is gene-level co-linearity (MCScanX, JCVI), whole-genome structural rearrangements (SyRI, AnchorWave), multi-genome macrosynteny (GENESPACE, ntSynt), or synteny-aware orthology (GENESPACE, ProteinOrtho-synteny). Repeat-masking quality is the dominant determinant of result reliability -- unmasked TEs produce ~100x more false anchor pairs than real syntenic anchors.

• CLI: MCScanX for collinear gene blocks via dynamic programming
• CLI: python -m jcvi.compara.catalog ortholog A B for JCVI/MCScan Python pipeline
• R: run_genespace() (Lovell 2022) for orthology-anchored riparian plots + pan-gene tracks
• CLI: syri for inversion / translocation / duplication detection
• CLI: anchorwave proali for sequence-level WGD-aware synteny

Algorithmic Taxonomy

Methodology evolves; GENESPACE has emerged as the de facto standard for plant comparative genomics (2022-2026). For non-plant clades, MCScanX or JCVI remain the workhorses. Whole-genome alignment for synteny is increasingly delegated to Cactus / Minigraph-Cactus and HAL toolkit (see [[whole-genome-alignment]]).

Decision Tree by Experimental Scenario

Per-Tool Failure Modes

Repeat-derived false synteny

Trigger: Running BLAST input for MCScanX/JCVI without softmasking repeats; or weak softmasking.

Mechanism: Transposable elements occupy 30-80% of plant and animal genomes; unmasked TEs produce millions of paralogous BLAST hits forming "synteny" blocks that are TE-driven, not real ancestry. The dynamic programming sees long apparent collinear blocks of TE-encoded proteins.

Symptom: MCScanX/JCVI output dominated by short (5-10 anchor) blocks; blocks cluster in TE-rich regions (heterochromatin, centromeric); BLAST hit count > 10x what is expected for the divergence.

Fix: Softmask genomes with RepeatModeler2 (de novo TE library) + RepeatMasker before BLAST; report softmasking statistics. For MCScanX, set -e 1e-10 (stricter) for close species, -e 1e-5 for distant. JCVI --cscore 0.95 enforces near-reciprocal-best-hit, dramatically reducing TE-driven blocks. Always check that block size distribution drops at 5-10 anchors (TE-driven) vs proper > 20 anchor blocks (real synteny).

Reference-guided assembly creating circular synteny

Trigger: Comparing a de novo assembly with a reference-guided scaffold of the same / related species.

Mechanism: Reference-guided scaffolding orders contigs based on synteny to the reference. Comparing the resulting scaffold to that same reference produces "synteny" that is actually the imposed reference order, not biological gene order. The comparison is circular.

Symptom: Suspicious uniformity in synteny across most chromosomes; visualization shows reference and query as perfectly parallel; SyRI reports few SVs where divergence-time-appropriate counts would predict many.

Fix: Verify assembly method via metadata. For reference-guided assemblies, treat synteny inferences with the reference as suspect; use a sister species reference or pure de novo assembly. Hi-C-scaffolded assemblies are not "reference-guided" in this sense and are safe.

Fragmented assemblies underestimating macrosynteny

Trigger: Running synteny on draft assemblies with N50 < 1 Mb.

Mechanism: Macrosynteny detection requires that syntenic blocks be on single contigs; if contigs are short, blocks are artificially truncated. JCVI reports show artifactual "synteny loss" between species when one assembly is fragmented.

Symptom: Synteny "loss" correlates with assembly N50; per-chromosome dotplots show dotted lines where solid diagonals expected; SyRI reports excessive INVs / TRANS where the divergence is actually low.

Fix: Require minimum N50 of 1 Mb for cross-species synteny; chromosome-level assemblies preferred for SyRI. For fragmented assemblies, restrict analysis to long-contig regions or use SyRI's contig-level mode with caveats.

Tandem duplicate inflation

Trigger: Including tandemly duplicated genes in synteny anchor input.

Mechanism: Tandem duplicates produce many same-region BLAST hits between species; the dynamic programming includes them as multiple anchors per gene position, artificially inflating block size and confidence.

Symptom: "High-confidence" synteny blocks in regions known for tandem expansions (NLR clusters in plants, OR genes in mammals); blocks anchored largely on adjacent tandem duplicates of the same gene.

Fix: MCScanX automatically collapses tandem duplicates (within 5 genes by default; -w parameter). Verify in the output the tandem-collapse step ran. JCVI --tandem_Nmax 10 is more aggressive. For NLR / OR genes specifically, manual tandem filtering before BLAST is preferred.

Cross-genome chromosome name mismatch

Trigger: GFF and FASTA from different sources with different chromosome naming (chr1 vs Chr1 vs 1 vs NC_001234.5).

Mechanism: MCScanX requires chromosomes to be named consistently; mismatch causes empty output or partial blocks.

Symptom: MCScanX completes but .collinearity file is small or empty; warnings about unknown chromosome IDs.

Fix: Normalize chromosome names with awk or samtools faidx --regions to unify naming convention; verify GFF + FASTA share the same names. For NCBI assemblies, use datasets to download with consistent labels.

SyRI complaining about inversions in absence of true inversions

Trigger: Running SyRI on alignments with many small-scale rearrangements; or when one genome has been re-ordered (different convention for which strand is "+").

Mechanism: SyRI's syntenic-path algorithm penalizes any rearrangement; high-divergence pairs accumulate small-scale rearrangements that are not "real" inversions but inheritance differences.

Symptom: SyRI reports thousands of small "inversions" (< 1 kb); blocks dominate the SV count; total INV length unrealistic.

Fix: Filter SyRI output by size: real biologically-relevant INVs are typically > 5 kb. Tighten minimap2 sensitivity (-x asm5 for close species; -x asm10 or -x asm20 for more divergent). Reverse-complement one assembly's chromosomes if the strand convention differs.

GENESPACE OrthoFinder version mismatch

Trigger: Running GENESPACE with an outdated bundled OrthoFinder.

Mechanism: GENESPACE bundles a specific OrthoFinder version; if user's installed OrthoFinder is newer (v3 vs v2.5), the HOG output layout differs and GENESPACE parsing fails.

Symptom: GENESPACE reports "Phylogenetic_Hierarchical_Orthogroups not found" or similar; orthogroup table is empty.

Fix: Use the OrthoFinder version GENESPACE expects (currently OrthoFinder 2.5.4 for GENESPACE 1.4.x). Pin via conda env. Future GENESPACE versions will adapt to OrthoFinder 3 layout (Lovell update expected).

Microsynteny vs macrosynteny conflation

Trigger: Reporting "synteny" without specifying scale.

Mechanism: Macrosynteny (same chromosome) is detectable across hundreds of millions of years but decays as rearrangements accumulate; microsynteny (same order within local region) can be deeply conserved even when macrosynteny is lost. Conflating them produces misleading "synteny loss" claims.

Symptom: Reports of "synteny lost between species X and Y" when both share microsyntenic gene clusters; or vice versa.

Fix: Always specify scale. SynNet (Zhao 2017) explicitly separates the two. MCScanX -s parameter (minimum block anchors) at 5 = microsynteny; at 20+ = macrosynteny. Report block-size distribution.

Polyploid / WGD-affected genome ambiguity

Trigger: Running MCScanX on a polyploid (e.g. allohexaploid wheat) without subgenome assignment.

Mechanism: WGD doubles all genes; synteny is between subgenomes within the polyploid AND across to outgroup. Without subgenome assignment, all paralogs and orthologs collapse into the same blocks.

Symptom: Many "1:many" or "many:many" synteny relationships; per-chromosome synteny counts indicate doubled or tripled blocks; Ks distribution multimodal.

Fix: Use AnchorWave proali with explicit ploidy specification. Alternatively, run synteny twice: within polyploid (subgenome-vs-subgenome) and across to outgroup. See [[whole-genome-duplication]] for subgenome assignment workflow.