Computational protocol: A classification model to predict synergism/antagonism of cytotoxic mixtures using protein-drug docking scores

Similar protocols

Protocol publication

[…] The MixLow method [] (and Boik J, Narasimhan B: Introducing the R package mixlow for assessment of drug synergism/antagonism, submitted) was used to assess drug interactions based on results from the cytotoxicity assays. In brief, the MixLow method utilizes a nonlinear mixed-effects model to accurately estimate parameters of sigmoidal concentration-effect curves. The parameter estimates are used to construct an estimator of a Loewe-additivity interaction index, Lˆϕ. Index estimates are obtained at various fraction affected (φ) values, where fraction affected refers to the expected fraction of the cell population affected by a given drug concentration. An interaction index of 1.0 at a specified fraction affected indicates additivity, an index less than 1.0 indicates synergism, and an index greater than 1.0 indicates antagonism. A short mathematical explanation of the MixLow method and the Loewe index is provided in Additional File .An example of fraction affects vs. estimated interaction index is given in Figure for the mixture M15. Ninety-five percent confidence intervals of the index are also shown (dotted lines). Statistically significant synergism is indicated for this mixture between 0.10 ≤ φ ≤ 0.85 (where both confidence intervals are less than 1.0) and antagonism is indicated at a fraction affected greater than about 0.93 (where both intervals are greater than 1.0). Additivity is indicated at a fraction affected less than 0.1 (where the confidence intervals span 1.0).The responses (synergism scores) modeled in this paper were constructed as an area under the curve of statistically significant antagonism minus the area under the curve of statistically significant synergism. Index values over the interval 0.31 ≤ φ ≤ 0.64 were used. For example, in Figure both confidence intervals are below 1.0 in this interval. Thus the area under the curve of statistically significant antagonism is zero and that for statistically significant synergism is the area between the upper confidence interval and the reference line of 1.0 over this φ interval. The calculated response would thus be negative, indicating that on average there was more synergism than antagonism occurring. A calculated response of zero would indicate that no significant synergism or antagonism is occurring (i.e., additivity cannot be ruled out over this φ interval) or that antagonism and synergism are occurring but are balanced. The interval 0.31 ≤ φ ≤ 0.64 was used because lower values of φ would not have a large impact on cell viability and higher values of φ may require excessive drug concentrations (greater than might be achieved clinically). Moreover, confidence intervals are typically wider at extreme φ values. Preliminary models that used the interaction index itself as the response variable were less accurate than ones based on confidence intervals (data not shown). Programming of the MixLow method was done in the R [] and Python (with Numpy/Scipy) [,] environments. […]

Pipeline specifications

Software tools Numpy, SciPy
Applications Miscellaneous, WGS analysis
Organisms Homo sapiens
Diseases Lung Neoplasms, Drug-Related Side Effects and Adverse Reactions
Chemicals Curcumin, Doxorubicin, Quercetin