SeqScreen: accurate and sensitive functional screening of pathogenic sequences via ensemble learning

Area of Science:

• Microbiology
• Bioinformatics
• Genomics

Background:

• The COVID-19 pandemic highlighted the critical need for precise pathogen detection.
• Characterizing pathogenic genetic sequences remains a significant scientific challenge.
• Existing methods may lack the specificity required for identifying novel or concerning sequences.

Purpose of the Study:

• To develop a computational tool, SeqScreen, for accurate characterization of short nucleotide sequences.
• To implement a machine learning model for assigning taxonomic and functional labels to sequences.
• To create a specialized database of Functions of Sequences of Concern (FunSoCs) relevant to microbial pathogenesis.

Main Methods:

• Development of an ensemble machine learning model.
• Utilizing a custom-curated set of FunSoCs for microbial pathogenesis.
• Training the model to label protein-coding sequences with FunSoCs.
• Validation of the model's performance using recall and precision metrics.

Main Results:

• SeqScreen demonstrates high recall and precision in labeling protein-coding sequences with FunSoCs.
• The tool effectively characterizes short nucleotide sequences with both taxonomic and functional information.
• Successful application of machine learning for pathogen-related sequence annotation.

Conclusions:

• SeqScreen represents a significant advancement in pathogen characterization and synthetic DNA screening.
• The approach offers a functionally informed method for analyzing genetic sequences.
• SeqScreen is available as an open-source tool to facilitate research in microbial pathogenesis and biosecurity.