Raktim Mitra led this work in which we used recurrent variational autoencoders to model gene expression dynamics from time series data. We can accurately predict gene dynamics from both bulk and single-cell gene expression datasets. Here is the preprint, open-source code, and explainer on twitter.