FURI | Spring 2026

Building a Machine Learning Model of Murine Vaccine Response

Flow Cytometry is a laboratory technique that measures fluorochrome intensities of individual cells to detect the presence of certain antigens. Upon processing these data files to obtain cell population percentages, the Machine Learning algorithms that map these input features to outputs must be trained and validated. This is a challenge because the data sets produced by Flow Cytometry are very voluminous, and labels, or outputs, for the training data can be difficult to define. In this research, methods for generating training data, such as bootstrapping and concatenating data from different vaccine treatment groups, are investigated. The primary strategy in this research is to automate a kernel function search, employ sequential feature selection algorithms to identify the most decisive cell populations, and use vaccine schedule groups as labels. If this automated kernel search returns high accuracy scores on training data, it will enable medical experts to make conclusions about vaccine efficacy much more quickly.