Machine learning regression based quantification of dynamic movements of prokaryotic bacterial type IV pili.

Abstract

The surfaces of many bacteria contain long dynamic appendages called type IV pili (T4P) Researchers have determined the correlation between the movements of T4P and the virulence of bacterial cell colonies. Previously, these movements were quantified through manual hand count- ing systems – a tedious process, taking researchers hours to multiple days to process one colony. This study aimed to develop an automated program to quantify the dynamic movements of T4P in bacterial samples using regression techniques and machine learning. A simple random sampling approach evaluated the program’s accuracy by comparing its output to manual hand-counted data. The results showed an average accuracy of 91.11% across ten analyzed files, with a standard error of 0.43%. This study demonstrates the potential of automated image analysis techniques to expedite the quantification of complex bacterial behaviors, such as T4P dynamics while maintaining high accuracy.

Keywords

• prokaryotic bacteria
• type iv pili
• elliptic regression
• transfer machine learning