Reference: Recursive model for dose-time responses in pharmacological studies
Majority of the drug sensitivity predictive modeling approaches attempt to predict a single representative metric of the complete dose-response curve such as the half maximal inhibitory concentration (IC50) or the area under the curve (AUC) at the steady-state of 72 hours. This can potentially fail to provide some crucial information such as the trend of change in sensitivity as dose increases or the difference in sensitivity trends between two dose-response curves with similar AUC and/or IC50 values. These information can often be used for a particular patient receiving precision therapy to select the most effective drug dosage or to avoid potential drug toxicity at the current time point.
We use in vitro dose-time proteomic and cellular viability information post drug administration for 10 BRAF Melanoma V600E/D cell lines from the HMS-LINCS database which is part of the NIH Library of Integrated Network-based Cellular Signatures (LINCS) Program at Harvard Medical School. We were forced to limit our analysis to a single dataset since, to our knowledge, HMS-LINCS is the only publicly available source offering functional responses as well as predictors. To know more about the data, please look into our investigation in modeling drug sensitivity using proteomic features in our previous work from BHI 2018: An investigation of proteomic data for application in precision medicine.
We incorporate two well-established models to describe the cell behavior in dose and time post drug administration, therefore, developing a joint dose-time response model. For individual dose-response curves at a certain time point, we use the Hill equation or the 4-parameter sigmoidal model while to explain the temporal behavior of such curves, we apply Gompertz law, a renowned tumor growth modeling approach, to explain cellular viability in time. The amalgamation of these two models yields the following joint model for explaining drug sensitivity at any dose-time point -
From this equation, we can get a recursive relation for drug sensitivity between two immediate time points t and t-, thus, we can perform a one-step prediction using the temporal trend in proteomic data at time point t- post drug administration.
The details of the model is described in our 2019 paper: Recursive model for dose-time responses in pharmacological studies.
This repository contains the necessary code to reproduce the results described in the paper and the corresponding source for the data used in the simulation experiments.
- Data: Contains data for modeling. The processed data used in our experiments can be found here
- Data/
synthetic_recursive.mat: Data file for synthetic experiment RecursiveHybridModel.m: Model class with all necessary methodssynthetic_data_analysis_SRD_v2.m: Synthetic experiment code for modeling synthetic dataLINCS_data_preprocessing.m: Code for preprocessing HMS-LINCS proteomics dataLincs_72hr_impute.m: Code for imputing proteomic expression values at 72 hours using spline interpolationlincs_rf_dataset_SRD_v2.m: Code for modeling using HMS-LINCS data
If you use our Recursive Hybrid approach for your research/application, please cite the following paper -
Dhruba, S.R., Rahman, A. et al. Recursive model for dose-time responses in pharmacological studies. BMC Bioinformatics 20, 317 (2019). DOI: https://doi.org/10.1186/s12859-019-2831-4
If you use our work on the exploration of proteomic data for predictive modeling of drug sensitivity, please cite the following paper -
Matlock, K., Dhruba, S. R. et al., An investigation of proteomic data for application in precision medicine, 2018 IEEE EMBS International Conference on Biomedical & Health Informatics (BHI), Las Vegas, NV, 377-380 (2018). DOI: https://doi.org/10.1109/BHI.2018.8333447