This repository contains a jupyter notebook containing the calculations described in the paper Metabolic imaging across scales reveals distinct prostate cancer phenotypes.
In the notebook we analyse mass spectrometry imaging (MSI) data from a set of prostate cancer (PCa) samples.
We train a small neural network to classify samples as either bengign or malignant. We use Shapley additive explanations to investigate the contributions of each metabolite to a sample's classification as either malignant or bengign.