Application of transfer learning for cancer drug sensitivity prediction

Abstract

BackgroundIn precision medicine, scarcity of suitable biological data often hinders the design of an appropriate predictive model. In this regard, large scale pharmacogenomics studies, like CCLE and GDSC hold the promise to mitigate the issue. However, one cannot directly employ data from multiple sources together due to the existing distribution shift in data. One way to solve this problem is to utilize the transfer learning methodologies tailored to fit in this specific context.ResultsIn this paper, we present two novel approaches for incorporating information from a secondary database for improving the prediction in a target database. The first approach is based on latent variable cost optimization and the second approach considers polynomial mapping between the two databases. Utilizing CCLE and GDSC databases, we illustrate that the proposed approaches accomplish a better prediction of drug sensitivities for different scenarios as compared to the existing approaches.ConclusionWe have compared the performance of the proposed predictive models with database-specific individual models as well as existing transfer learning approaches. We note that our proposed approaches exhibit superior performance compared to the abovementioned alternative techniques for predicting sensitivity for different anti-cancer compounds, particularly the nonlinear mapping model shows the best overall performance.Electronic supplementary materialThe online version of this article (10.1186/s12859-018-2465-y) contains supplementary material, which is available to authorized users.

From Medical Imaging to Radiomics: Role of Data Science for Advancing Precision Health

“…From a modeling standpoint, two main developments are currently extremely useful: transfer learning (TL) [36][37][38][39][40][41][42][43] and ensemble modeling (EM) approaches (see general introduction to the topic in [44]). TL assumes that the features learned in a certain application domain can be usefully applied to a different related domain.…”

The International Conference on Intelligent Biology and Medicine (ICIBM) 2018: bioinformatics towards translational applications

“…However, integrating information from multiple resources faces the challenge of removing the distribution shift between data. Dhruba et al [5] proposed to use transfer learning methodologies to eliminate this distribution shift and design effective drug sensitivity prediction models in a target database by incorporating data from a secondary database. More specifically, the authors presented two novel approaches based on latent variable cost optimization and polynomial mapping.…”

Ensemble transfer learning for the prediction of anti-cancer drug response

“…In the context of drug response prediction, the target and source domains of transfer learning can be different drug screening studies/datasets 39 . Differences in experimental protocols, assays, or biological models and drugs used in the studies generate variations between these datasets.…”

“…It has been reported that the same treatment experiments (i.e. pairs of drugs and CCLs) might have quite different response values in different studies 39 . Supplementary Fig.…”

“…Compared with the above existing works, our work makes unique contributions. First, while existing works on transfer learning for drug response prediction focus on building prediction models for a specific drug [39][40][41][42] , we target the more challenging task of building general prediction models that are not specific to a drug. Different from drug-specific prediction models, general drug response prediction models are trained on data of multiple drugs.…”