Learning Individual Survival Models from PanCancer Whole Transcriptome Data

Personalized medicine attempts to predict survival time for each patient, based on their individual tumour molecular profile. We investigate whether our survival learner in combination with a dimension reduction method can produce useful survival estimates for a variety of cancer patients. This paper provides a method that learns a model for predicting the survival time for individual cancer patients from the PanCancer Atlas: given the (16335-dimensional) gene expression profiles from 10173 patients, each having one of 33 cancers, this method uses unsupervised NMF (non-negative matrix factorization) to re-express the gene expression data for each patient in terms of 100 learned NMF Factors. It then feeds these 100 Factors into the MTLR (Multi-Task Logistic Regression) learner to produce cancer-specific models for each of 20 cancers (with >50 uncensored instances); this produces "individual survival distributions" (ISDs), which provide survival probabilities at each future time for each individual patient - which provides a patient's risk score and estimated survival time. Our NMF-MTLR concordance indices outperformed the VAECox benchmark by 14.9% overall. We achieved optimal survival prediction using pan-cancer NMF in combination with cancer-specific MTLR models. We provide biological interpretation of the NMF model and clinical implications of ISDs for prognosis and therapeutic response prediction. NMF-MTLR provides many benefits over other models: superior model discrimination, superior calibration, meaningful survival time estimates, and accurate probabilistic estimates of survival over time for each individual patient. We advocate for the adoption of these cancer survival models in clinical and research settings.