OGGfinder: Accurate Orthogroup Inference for Pan-Gene Families in Complex Genomes

Abstract

Accurate inference of orthologous gene groups (OGGs) is a foundational step in comparative genomics, yet existing tools fail to meet the demands of complex allopolyploid genomes. Here, we present OGGfinder, a novel pipeline that integrates sequence similarity with phylogenetic tree topology constraints, a data-driven 5th percentile (P5) threshold inference mechanism, a robust 6-step post-processing pipeline, and Latin Hypercube Sampling (LHS) for automated parameter optimization. In a benchmark utilizing 2,920 AP2 gene family from 164 allopolyploid cotton (Gossypium) genomes with a target orthogroup size of 164 genes, OGGfinder successfully recovered 18 high-quality OGGs with a mean size of 162.2 genes and zero singletons, tightly approximating the expected species count. In contrast, OrthoFinder drastically over-clustered genes into only 7 massive groups (mean size 417.1, max 654), while TreeCluster heavily fragmented the data into 116 groups with 42 singletons (36.2% singleton rate). CD-HIT generated 22 groups with a median size of only 52.5 and discarded nearly 1,000 sequences (34.6% gene loss) due to its greedy redundancy-reduction strategy. Comprehensive six-dimensional evaluation (completeness, granularity, topology consistency, auto-parameterization, polyploidy support, and scalability) yielded total scores of OGGfinder 26.0, OrthoFinder 23.9, CD-HIT 20.0, and TreeCluster 13.7 out of 30. These results demonstrate that OGGfinder significantly outperforms existing state-of-the-art tools, offering a highly accurate and reproducible solution for pan-gene family analyses in polyploid species. This is particularly critical for the application of finding OGGs within pan-gene families.