MFmap: A semi-supervised generative model matching cell lines to tumours and cancer subtypes

Abstract

Translating in vitro results from experiments with cancer cell lines to clinical applications requires the selection of appropriate cell line models. Here we present MFmap (model fidelity map), a machine learning model to simultaneously predict the cancer subtype of a cell line and its similarity to an individual tumour sample. The MFmap is a semi-supervised generative model, which compresses high dimensional gene expression, copy number variation and mutation data into cancer subtype informed low dimensional latent representations. The accuracy (test set F1 score > 90%) of the MFmap subtype prediction is validated in ten different cancer datasets. We use breast cancer and glioblastoma cohorts as examples to show how subtype specific drug sensitivity can be translated to individual tumour samples. The low dimensional latent representations extracted by MFmap explain known and novel subtype specific features and enable the analysis of cell-state transformations between different subtypes. From a methodological perspective, we report that MFmap is a semi-supervised method which simultaneously achieves good generative and predictive performance and thus opens opportunities in other areas of computational biology. Author summaryCancer researchers perform experiments with cell lines to better understand the biology of cancer and to develop new anti-cancer treatments. A prerequisite to translate promising results from these in vitro experiments to clinical applications is to use the most appropriate cell line for a given tumour or cancer subtype. We present MFmap (model fidelity map), a deep learning technique to integrate cancer genomic data from patients with cell line data. The MFmap neural network compresses complex genomic features from thousands of genes into a small set of features called latent representations. This makes cell line and tumour data comparable and allows cancer researchers to select the best cell line which closely resembles a specific type of tumours or even an individual tumour. By classifying cancer cell lines into subtypes, MFmap offers a new possibility to predict the effect of therapeutic compounds in a particular tumour subtype. For the example of an aggressive brain tumour we demonstrate that MFmap can be used to study cell-state transformations during the disease course. In addition, MFmap is a promising machine learning method with potential applications in many other areas of biology and medicine.