Systematic assessment of multi-gene predictors of pan-cancer cell line sensitivity to drugs exploiting gene expression data

Abstract

Background: Selected gene mutations are routinely used to guide the selection of cancer drugs for a given patient tumour. Large pharmacogenomic data sets were introduced to discover more of these single-gene markers of drug sensitivity. Very recently, machine learning regression has been used to investigate how well cancer cell line sensitivity to drugs is predicted depending on the type of molecular profile. The latter has revealed that gene expression data is the most predictive profile in the pan-cancer setting. However, no study to date has exploited GDSC data to systematically compare the performance of machine learning models based on multi-gene expression data against that of widely-used single-gene markers based on genomics data. Methods: Here we present this systematic comparison using Random Forest (RF) classifiers exploiting the expression levels of 13,321 genes and an average of 501 tested cell lines per drug. To account for time-dependent batch effects in IC 50 measurements, we employ independent test sets generated with more recent GDSC data than that used to train the predictors and show that this is a more realistic validation than K-fold cross-validation. Results and Discussion: Across 127 GDSC drugs, our results show that the single-gene markers unveiled by the MANOVA analysis tend to achieve higher precision than these RF-based multi-gene models, at the cost of generally having a poor recall (i.e. correctly detecting only a small part of the cell lines sensitive to the drug). Regarding overall classification performance, about two thirds of the drugs are better predicted by multi-gene RF classifiers. Among the drugs with the most predictive of these models, we found pyrimethamine, sunitinib and 17-AAG. Conclusions: We now know that this type of models can predict in vitro tumour response to these drugs. These models can thus be further investigated on in vivo tumour models.

Predicting Cancer Drug Response In Vivo by Learning an Optimal Feature Selection of Tumour Molecular Profiles

“…This means that many responders without the marker are now correctly identified as such, owing to an effective combination of multiple gene alterations. In other words, multi-gene predictors of in vivo drug response generally have a higher recall than single-gene markers, which confirms previous findings in vitro [13,43,45]. The recall of a single-gene marker will be necessarily poor in all cases in which the prevalence of the mutation is much lower than the response rate.…”

“…One study [69] applied several ML algorithms to predict pancancer cell line response from transcriptomic profiles, obtaining MCCs below 0.6 in all cases (see Figure 1 in that paper). Maximum MCCs slightly above 0.5 and 0.3 were also obtained using RF with transcriptomic profiles [43] and genomic profiles [13], respectively. Another study [70] also predicted drug response using many hundreds of pancancer cell lines and several ML algorithms from various omics profiles (gene expression, copynumber alterations, single-nucleotide mutations).…”

“…While typically, only tens of tumours have their response to the drug available, the molecular profiles of these tumours may easily aggregate over 50,000 features. To face this challenge, ML algorithms with built-in FS, such as Elastic Nets [36][37][38][39][40], Ridge [37,40], LASSO [37,38,40,41], Random Forest (RF) [13,37,39,40,42,43] or XGBoost [44,45], have been used to model pharmacogenomics data from in vitro cell lines. However, these methods have also been found to be unable to provide predictive models for many drugs.…”