CSS
We proposed a novel experimental design to enable a more cost-effective testing of synergy and sensitivity for a drug pair. First, the dose-response curve for each single drug is determined. Then the drug at its IC50 concentration is combined with the other drug at multiple doses, generating an IC50-based dose-response curve for the drug pair. We developed a drug combination sensitivi-ty score (CSS) to summarize the dose-response curves. Using high-throughput drug combination data from cancer cell lines, we showed that the CSS is highly reproducible among the replicates. With machine learning approaches such as elastic net, random forests and support vector ma-chines, the CSS can also be predicted at high accuracy using features including drug-target in-teraction profiles and structural fingerprints. Furthermore, we derived a synergy score based on the difference between the drug combination and the single drug dose-response curves. We showed that the CSS-based synergy score is able to detect true synergistic and antagonistic drug combinations. The IC50-based experimental design coupled with the CSS scoring facilitated the evaluation of drug combination sensitivity and synergy using the same unit, with minimal ex-perimental material that is required to achieve sufficient prediction accuracy. The experimental and computational strategy could be utilized as an efficient platform for improving the discovery rate in high-throughput drug combination screening.
Install / Use
/learn @amalyutina/CSSREADME
R scripts:
-
calculate_sensitivity.R This is the last version of CSS calculation for drug combinations and sensitivity (relative inhibition - RI) calculation for single agents. Datasets, such as single_drug_data.RData and combo_drug_data.RData can be utilized to calculate those scores.
-
predicion_of_css.Rmd The script predicts the CSS for drug combinations from css_data.RData using drug target information from fingerprints.RData, primary_targets.RData and primary_plus_sea_targets.RData, and machine learning methods (Elastic Net, Random Forests, Support Vector Machines). R2, RMSE, MAE and correlation values are calculated for a selected cell line and machine learning method.
RData files:
-
single_drug_data.RData A file that can be used in calculate_sensitivity.R to calculate relative inhibition (RI) for single drug experiments. The file has to contain such columns as cellLine, drug, dose and viability.
-
combo_drug_data.RData A file that can be used in calculate_sensitivity.R to calculate CSS for combination drug experiments. The file has to contain such columns as cellLine, drugA, drugAdose, drugB, drugBdose and viability.
-
fingerprints.RData A file that can be used in predicion_of_css.Rmd to build a matrix of features. It contains binary fingerprint data mapped to the drugs used in the experiment. It has to contain column with drug names named ‘drug’.
-
primary_targets.RData A file that can be used in predicion_of_css.Rmd to build a matrix of features. It contains binary primary drug target data mapped to the drugs used in the experiment. It has to contain column with drug names named ‘drug’.
-
primary_plus_sea_targets.RData A file that can be used in predicion_of_css.Rmd to build a matrix of features. It contains binary primary drug target data and drug targets that were identified using SEA method mapped to the drugs used in the experiment. It has to contain column with drug names named ‘drug’.
-
css_data.RData A file that can be used in predicion_of_css.Rmd. The file provides CSS data that were calculated using calculation_of_css.Rmd. It has to contain columns with both drug names named ‘drug1’ and ‘drug2’, cell lines specified in column ’cellLine’ and CSS column names ‘CSS’.
