Pre-Training on In Vitro and Fine-Tuning on Patient-Derived Data Improves Deep Neural Networks for Anti-Cancer Drug-Sensitivity Prediction

SIMPLE SUMMARY: Cancer cell lines vary greatly from one another because each one underwent a combination of random mutations. Therefore, the effectiveness of anti-cancer drugs also varies across cancer cell lines. In order to understand which drugs are effective against which cancer cells, researchers expose cultivated cell lines to drug candidates. However, these results are not entirely realistic, mostly because cultivated cells live in a two-dimensional in vitro environment without interacting with other types of cells. There are more realistic ways to test drug effectiveness—e.g., organoids that are 3D-printed from tumor cells—but since these processes are more complex, much fewer data are available. We studied an approach in which a neural network is first trained on the wealth of available in vitro drug-sensitivity data, before being fine-tuned on smaller but more realistic databases. We found that this training procedure improves the neural network’s ability to predict how effective a particular drug will be for a given tumor cell line. Such neural networks can serve as a tool, both for personalized treatment of cancer and for drug development. ABSTRACT: Large-scale databases that report the inhibitory capacities of many combinations of candidate drug compounds and cultivated cancer cell lines have driven the development of preclinical drug-sensitivity models based on machine learning. However, cultivated cell lines have devolved from human cancer cells over years or even decades under selective pressure in culture conditions. Moreover, models that have been trained on in vitro data cannot account for interactions with other types of cells. Drug-response data that are based on patient-derived cell cultures, xenografts, and organoids, on the other hand, are not available in the quantities that are needed to train high-capacity machine-learning models. We found that pre-training deep neural network models of drug sensitivity on in vitro drug-sensitivity databases before fine-tuning the model parameters on patient-derived data improves the models’ accuracy and improves the biological plausibility of the features, compared to training only on patient-derived data. From our experiments, we can conclude that pre-trained models outperform models that have been trained on the target domains in the vast majority of cases.