Seqping: gene prediction pipeline for plant genomes using self-training gene models and transcriptomic data.

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Accurate gene prediction is crucial for genome annotation, yet current tools struggle with comprehensive protein-coding region identification.
• Existing gene-finders lack universal Hidden Markov Models (HMMs) for effective, automatic gene prediction across all organisms.

Purpose of the Study:

• To develop an automated gene prediction pipeline, Seqping, that generates species-specific HMMs for unbiased predictions.
• To improve the accuracy and efficiency of gene prediction in newly sequenced or understudied genomes.

Main Methods:

• Seqping integrates GlimmerHMM, SNAP, and AUGUSTUS pipelines with MAKER2 for combining predictions.
• Utilizes self-training HMMs and transcriptomic data for species-specific model generation.
• Employs Benchmarking Universal Single-Copy Orthologs (BUSCO) for pipeline evaluation.

Main Results:

• Seqping achieved high accuracy in gene prediction for *Oryza sativa* (rice) and *Arabidopsis thaliana*.
• The pipeline identified at least 95% of the BUSCO plantae dataset, demonstrating broad applicability.
• Seqping outperformed MAKER2, GlimmerHMM, and AUGUSTUS in gene prediction accuracy using their default HMMs.

Conclusions:

• Seqping offers a seamless pipeline for training species-specific HMMs, enhancing gene prediction in novel genomes.
• The Seqping pipeline provides more accurate gene predictions compared to existing HMM-based approaches.