Heterogeneity Aware Random Forest for Drug Sensitivity Prediction - PubMed (original) (raw)

Heterogeneity Aware Random Forest for Drug Sensitivity Prediction

Raziur Rahman et al. Sci Rep. 2017.

Abstract

Samples collected in pharmacogenomics databases typically belong to various cancer types. For designing a drug sensitivity predictive model from such a database, a natural question arises whether a model trained on diverse inter-tumor heterogeneous samples will perform similar to a predictive model that takes into consideration the heterogeneity of the samples in model training and prediction. We explore this hypothesis and observe that ensemble model predictions obtained when cancer type is known out-perform predictions when that information is withheld even when the samples sizes for the former is considerably lower than the combined sample size. To incorporate the heterogeneity idea in the commonly used ensemble based predictive model of Random Forests, we propose Heterogeneity Aware Random Forests (HARF) that assigns weights to the trees based on the category of the sample. We treat heterogeneity as a latent class allocation problem and present a covariate free class allocation approach based on the distribution of leaf nodes of the model ensemble. Applications on CCLE and GDSC databases show that HARF outperforms traditional Random Forest when the average drug responses of cancer types are different.

PubMed Disclaimer

Conflict of interest statement

The authors declare that they have no competing interests.

Figures

Figure 1

Figure 1

Melanoma (Skin) tumorigenesis pathway, collected from KEGG.

Algorithm 1

Algorithm 1

Algorithmic representation of Heterogeneity Aware Random Forest (HARF) Regression.

Figure 2

Figure 2

3 sample trees with leaf information. Boxed numbers represent the samples contained within each leaf node. Red samples belong to cancer type C A while green samples belong to cancer type C B.

Figure 3

Figure 3

With an increase in the number of samples for training, the percentage of mis-classifications for HARF, Decision Tree and Linear Discriminant Analysis (LDA) all get reduced. Using drug Nilotinib of CCLE database and 2 cancer types HLT and Lung, this reduction of misclassification is shown. For small number of samples, HARF has the lowest misclassification rate. For large sample sizes, LDA gives the lowest misclassification rate, but the differences are minimal in both the cases.

Figure 4

Figure 4

Changes in misclassification rate of HARF and Bayes error (Eq. 16) for different number of trees are shown. For model with few trees, misclassification rate is higher compared to model with high number of trees. As expected, HARF misclassification rate is always higher compared to minimum Bayes error, but the difference is always minimal for models with different number of trees. Drug AZD − 6244 and cancer types Skin & CNS are used for the generation of these curves.

Similar articles

Cited by

References

    1. Zou H, Hastie T. Regularization and variable selection via the elastic net. Journal of the Royal Statistical Society: Series B (Statistical Methodology) 2005;67:301–320. doi: 10.1111/j.1467-9868.2005.00503.x. - DOI
    1. Barretina J, et al. The cancer cell line encyclopedia enables predictive modeling of anticancer drug sensitivity. Nature. 2012;483:603–607. doi: 10.1038/nature11003. - DOI - PMC - PubMed
    1. Gönen M, Margolin AA. Drug susceptibility prediction against a panel of drugs using kernelized bayesian multitask learning. Bioinformatics. 2014;30:i556–i563. doi: 10.1093/bioinformatics/btu464. - DOI - PMC - PubMed
    1. Costello JC, et al. A community effort to assess and improve drug sensitivity prediction algorithms. Nat. biotechnology. 2014;32:1202–1212. doi: 10.1038/nbt.2877. - DOI - PMC - PubMed
    1. Wan Q, Pal R. An ensemble based top performing approach for nci-dream drug sensitivity prediction challenge. PloS one. 2014;9:e101183. doi: 10.1371/journal.pone.0101183. - DOI - PMC - PubMed

Publication types

MeSH terms

LinkOut - more resources