Related Experiment Video
Updated: Aug 30, 2025
Evaluating the Effectiveness of Cancer Drug Sensitization In Vitro and In Vivo
Published on: February 6, 2015
Pre-Training on In Vitro and Fine-Tuning on Patient-Derived Data Improves Deep Neural Networks for Anti-Cancer
Paul Prasse1, Pascal Iversen1, Matthias Lienhard2
1Department of Computer Science, University of Potsdam, 14476 Potsdam, Germany.
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.
Insights
Pre-training machine learning models for drug sensitivity on in vitro data before fine-tuning with patient-derived data enhances accuracy. This approach improves preclinical drug development by leveraging existing data more effectively.
Area of Science:
- Computational biology
- Machine learning in oncology
- Drug discovery
Background:
- Machine learning models for drug sensitivity are trained on large in vitro cancer cell line databases.
- In vitro cell lines may not accurately represent human tumors due to long-term culture adaptation.
- Existing patient-derived data is insufficient for training high-capacity machine learning models.
Purpose of the Study:
- To improve the accuracy and biological relevance of machine learning models for predicting drug sensitivity.
- To investigate the efficacy of pre-training deep neural network models on in vitro data before fine-tuning on patient-derived data.
Main Methods:
- Utilized large-scale in vitro drug-sensitivity databases for pre-training deep neural network models.
- Fine-tuned pre-trained models using patient-derived drug-response data (cell cultures, xenografts, organoids).
- Compared the performance of pre-trained models against models trained solely on patient-derived data.
Main Results:
- Pre-training significantly improved the accuracy of drug-sensitivity prediction models.
- The biological plausibility of features learned by pre-trained models was enhanced.
- Pre-trained models demonstrated superior performance compared to models trained exclusively on patient-derived data in most cases.
Conclusions:
- Pre-training deep neural networks on in vitro drug-sensitivity data is a viable strategy to boost model performance.
- This approach addresses data scarcity issues with patient-derived datasets.
- Enhanced models offer improved accuracy and biological relevance for preclinical drug development.
