A computational approach to biological pathogenicity.

Area of Science:

• Microbiology
• Biochemistry
• Bioinformatics

Background:

• The COVID-19 pandemic highlighted the need for rapid methods to identify potentially pathogenic microbes.
• Existing methods for pathogen identification are often time-consuming and resource-intensive.
• A systematic approach to understanding pathogenicity is crucial for future pandemic preparedness.

Purpose of the Study:

• To develop a general, principled definition of pathogenicity.
• To create an operational framework for assessing microbial pathogenicity using DNA biomarkers.
• To predict the potential impact of microbes on a host, such as humans, a priori.

Main Methods:

• Utilizing basic biochemistry and Gibbs free energy of oligonucleotide duplex formation.
• Employing deep structural properties of DNA.
• Developing two operational tests based on the nearest neighbor (NN) model and the h-distance metric.
• Training machine learning models on DNA sequence features.

Main Results:

• The proposed tests achieved over 80% accuracy and over 90% sensitivity and specificity in predicting pathogenicity.
• Machine learning models demonstrated 90% accuracy, 100% sensitivity, and 80% specificity.
• These results indicate a high degree of reliability in predicting potential pathogens.

Conclusions:

• An objective, generalizable framework for pathogen identification based on DNA biomarkers is feasible.
• This approach can significantly reduce the need for costly and time-consuming clinical trials.
• The findings offer a transformative strategy for pandemic preparedness and mitigating adverse health impacts.