A variational deep-learning approach to modeling memory T cell dynamics.

After an influenza infection, T cells in the lung provide protection against future infection, but the concentration of such cells decays over time, leading to loss of immunity. These tissue-resident memory cells (TRM) are heterogeneous and not well characterized. To infer dynamical parameters from a timeseries of single-cell flow cytometry data, we develop a variational auto-encoder model that works for time-resolved data. The method can simultaneously clusters the data and learns dynamical parameters like loss rates and differentiation rates.

Recommended citation: van Dorp, CH. et al. (2025). "A variational deep-learning approach to modeling memory T cell dynamics." PLOS Computational Biology. 21(7): e1013242.
Download Paper