Predicting synergism of cancer drug combinations using NCI-ALMANAC data

Abstract

Background. Drug combinations are of great interest for cancer treatment. Unfortunately, the discovery of synergistic combinations by purely experimental means is only feasible on small sets of drugs. In silico modeling methods can substantially widen this search by providing tools able to predict which of all possible combinations in a large compound library are synergistic.Here we investigate to which extent drug combination synergy can be predicted by exploiting the largest available dataset to date (NCI-ALMANAC, with over 290,000 synergy determinations). Methods. Each cell line is modeled using primarily two machine learning techniques, RandomForest (RF) and Extreme Gradient Boosting (XGBoost), on the datasets provided by NCI-ALMANAC. This large-scale predictive modeling study comprises more than 5000 pair-wise drug combinations, 60 cell lines, 4 types of models and 5 types of chemical features. The application of a powerful, yet uncommonly used, RF-specific technique for reliability prediction is also investigated.Results. The evaluation of these models shows that it is possible to predict the synergy of unseen drug combinations with high accuracy (Pearson correlations between 0.43 and 0.86 depending on 2 the considered cell line, with XGBoost providing slightly better predictions than RF). We have also found that restricting to the most reliable synergy predictions results in at least two-fold error decrease with respect to employing the best learning algorithm without any reliability estimation. Alkylating agents, tyrosine kinase inhibitors and topoisomerase inhibitors are the drugs whose synergy with other partner drugs are better predicted by the models.Conclusions. Despite its leading size, NCI-ALMANAC comprises an extremely small part of all conceivable combinations. Given their accuracy and reliability estimation, the developed models should drastically reduce the number of required in vitro tests by predicting in silico which of the considered combinations are likely to be synergistic.

QSAR-derived affinity fingerprints (part 2): modeling performance for potency prediction

“…Overall, a correlation between predictive power and feature importance was not observed be the cause of this behaviour. To further evaluate the practical utility of QAFFP, future studies will be needed to challenge them in more complex scenarios, including the modeling of the synergistic or antagonistic effect of compound combinations [79][80][81][82], and to test whether the integration of QAFFP and cell line profiling data sets (e.g., basal gene expression profiles, or changes in gene expression induced upon compound administration) improves drug sensitivity modeling.…”

Concise Polygenic Models for Cancer-Specific Identification of Drug-Sensitive Tumors from Their Multi-Omics Profiles

“…However, even such learning algorithms tend to overfit training data due to their model complexity [ 18 ]. An overfitted model predicts training data much better than test data, which we have observed when using learning algorithms such as RF [ 15 , 19 ] or XGBoost [ 20 , 21 ]. An attractive way to reduce model complexity is reducing the number of input features, but if such a reduction is too large, then it will cause model underfitting (i.e., the model not being sufficiently complex for the data).…”

Essentiality and Transcriptome-Enriched Pathway Scores Predict Drug-Combination Synergy

“…Computational approaches to model and predict the synergy of drug combinations include machine-learning methods (DeepSynergy [ 15 ]; random forest (RF); extreme gradient boosting (XGBoost) [ 16 ]; and graph convolutional network (GCN) [ 17 ]), network methods [ 18 , 19 ], and systems biology methods [ 10 , 20 , 21 ]. Though these computational biology models differ in their analytical and theoretical methods, they share similar feature sets, including drug and cell-line features.…”