Pipeline Architecture & Benchmarks¶
This page summarizes the methodology and benchmark results reported in the orthoSynAssign preprint, orthoSynAssign: refine orthogroups using synteny information (Tsai, Pina Paez & Stajich, bioRxiv 2026).
Motivation¶
Clustering-based orthogroup inference (e.g. OrthoFinder) often produces complex, many-to-many orthogroups that mix true one-to-one orthologs with confounding paralogs. Incorporating synteny — the conserved local gene-order context around a gene — provides a robust way to split these clusters into high-granularity, single-copy syntenic orthogroups (SOGs). orthoSynAssign is a from-scratch rewrite of OrthoRefine, which pioneered this approach in C++ but suffered from sample-matching bugs and high memory usage at scale.
Hybrid Python/Rust Architecture¶
orthoSynAssign pairs an accessible Python command-line interface — for parsing OrthoFinder orthogroup files, BED annotations, and generating CLI/visualization output — with a synteny-comparison engine implemented in Rust. This split keeps installation and scripting simple (Python) while the performance-critical, per-gene neighborhood comparisons run in compiled, memory-efficient Rust code operating on optimized genomic-range data structures.
At a high level, for every gene shared between two genomes in a candidate orthogroup, orthoSynAssign inspects the -w/--window neighboring genes on either side and checks how many of them also have a syntenic (co-located) counterpart in the other genome. If the fraction of supporting neighbors meets the -r/--ratio_threshold, the gene pair is retained as syntenic, refining the original orthogroup into one or more smaller SOGs. See orthosynassign for the corresponding CLI options.
Multithreading Scalability Benchmark¶
To assess performance at scale, orthoSynAssign was benchmarked across 193 Eurotiomycetes genomes (a fungal class including Aspergillus and related genera) while varying the thread count (-t/--threads):
Runtime decreased substantially as thread count increased, confirming strong multithreaded scalability of the Rust synteny engine.
Memory usage remained strictly bounded and effectively thread-independent — i.e., running with more threads did not multiply peak memory usage, an important property for deployment on shared HPC systems with per-job memory limits.
These results indicate orthoSynAssign is well suited to high-throughput phylogenomic workflows involving hundreds of genomes.
See Also¶
orthosynassign and orthosynassign-vis command references