TGP-WEB: Trichodermaゲノムの自動遺伝子予測サーバー

Trichoderma Gene Prediction Web server (TGP-WEB) is designed for accurate gene prediction in genomes of Trichoderma species, the biological control agents and the plant-beneficial microorganisms. It employs a hybrid gene prediction strategy, combining ab initio (Augustus and GeneMark) and homology-based (Braker utilizing fungal protein sequences from NCBI Refseq database) methods. Predictions are integrated and prioritized using a ranking framework, followed by functional assessment via domain annotation. After uploading a single-genome FASTA file, users can download TGP-WEB prediction results (including GFF3 files with gene locations, and FASTA files for both the predicted nucleotide and amino acid sequences) within ~4 hours. TGP-WEB demonstrates high accuracy in gene prediction across 177 genomes. When benchmarked against 42 published genomes with annotated gene sets available on NCBI, it recovers more than 90.00% of reported genes in 37 (88.00%) genomes. For 135 previously unannotated genomes, TGP-WEB generates complete gene sets, now available on the web server. BUSCO evaluation shows greater than 97.00% completeness for 94.92% (168/177) of genomes. TGP-WEB predictions enable the identification of 2100 single-copy genes from Trichoderma genomes. These genes are used to construct a robust phylogenetic tree, which clarifies the taxonomy of 49 strains. The robust performance of TGP-WEB prediction will contribute to the increasing studies of Trichoderma genomes, and it is freely available from www.fungalgenomics.cn/geneprediction.