ProsmORF-pred: a machine learning-based method for the identification of small ORFs in prokaryotic genomes.

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Small open reading frames (smORFs) (<100 amino acids) are vital regulators in cellular processes.
• Computational identification of smORFs presents significant challenges.
• Accurate prediction of smORFs is crucial for understanding prokaryotic gene regulation.

Purpose of the Study:

• To develop a robust computational resource for predicting smORFs in prokaryotic genomes.
• To enhance the accuracy of smORF identification, particularly for short sequences (10-30 amino acids).
• To provide a tool for functional annotation of predicted smORFs.

Main Methods:

• Development of ProsmORF-pred, a machine learning-based prediction resource.
• Utilizing two ML models: one for initiation site recognition and another for protein-like sequence identification.
• Training models on known prokaryotic ORFs and annotated smORFs from Escherichia coli.
• Benchmarking against state-of-the-art tools on a dataset from 32 prokaryotic genomes.

Main Results:

• ProsmORF-pred demonstrates performance comparable to existing methods on general smORF sets.
• ProsmORF-pred significantly outperforms PRODIGAL and RANSEPS for predicting short smORFs (10-30 aa).
• The resource includes ProsmORFDB for functional annotation via sequence and genomic neighborhood similarity searches.

Conclusions:

• ProsmORF-pred offers a powerful and accurate solution for smORF identification in prokaryotes.
• The tool addresses a key challenge in predicting very short, functionally important smORFs.
• ProsmORF-pred is available as a user-friendly web server for broader accessibility.