Classification of T lymphocyte motility behaviors using a machine learning approach

Area of Science:

• Neuroimmunology
• Computational Biology
• Cellular Neuroscience

Background:

• T lymphocytes infiltrate organs, including the brain, contributing to neurological disorders.
• Understanding human T lymphocyte interactions with brain cells (astrocytes, neurons) is crucial but challenging.
• Manual analysis of T lymphocyte behavior is time-consuming and labor-intensive.

Purpose of the Study:

• To develop and validate a machine-learning model for automated classification of human CD8+ T lymphocyte behaviors.
• To analyze T lymphocyte interactions with human astrocytes and neurons in vitro.
• To reduce the time and labor required for characterizing T cell motility patterns.

Main Methods:

• Time-lapse microscopy of human CD8+ T lymphocytes co-cultured with primary human astrocytes or neurons.
• Training and validation of a machine-learning model (balanced random forest) for behavior classification.
• Feature selection using minimum redundancy maximum relevance algorithm during cross-validation.

Main Results:

• The machine-learning model achieved high accuracy in classifying T lymphocyte behaviors.
• Tested on astrocyte co-cultures, the model showed promising performance with a new experimenter.
• On an independent neuron co-culture dataset, the model achieved 0.82 binary and 0.79 3-class accuracy.

Conclusions:

• Automated classification of T lymphocyte behaviors using machine learning is feasible and efficient.
• This approach significantly reduces analysis time compared to manual methods.
• The model facilitates the study of T lymphocyte interactions with various organ-specific cells, aiding neurological research.