Using single-cell data from cancer samples presents a unique opportunity to understand patient outcomes. Data can be evaluated across scales: at the single-cell level, at the cell population level, at the patient level. We use this data to develop new algorithmic approaches to this type of data with a goal of informing patient outcomes in cancer. This involves machine learning, statistics and neural network models. We also focus on models that are human-interpretable and that provide insight into biology, as opposed to black boxes where the signals driving the results are obscure. We maintain access to these tools through our laboratory Github repository.