Automated Analysis of Complex Biphasic PROTAC Dose Response Data at Scale in Drug Development
A precise, quantitative and reproducible estimation of drug parameters is essential for robust Structure Activity Relationship (SAR) driven drug development. Traditional small molecule screening for inhibitors of the majority of targets yields monotonic dose response curves with a sigmoidal shape, characterised by a plateau at high drug concentrations. The compound is evaluated by fitting to a Hill model, leading to measurements of the maximum effect (efficacy) and the concentration for half maximal effect (IC50). In contrast, some drug modalities lead to a distinct dose response profile, marked by a loss of efficacy after the plateau. Such a “hook” effect is particularly relevant for PROteolysis TArgeting Chimeras (PROTACs). PROTACs are an exciting new modality that induce degradation rather than inhibition of targets, thus enabling the targeting of proteins without defined active sites such as scaffolding complexes. PROTACs induce ternary complex formation between the target, an endogenous E3 ligase and the PROTAC leading to protein degradation. Consequently, at high concentrations, PROTAC assays often exhibit a hook effect due to formation of independent binary complexes. Application of the standard Hill model to data with a hook effect results in misestimation of both IC50 and efficacy. Accounting for the hook effect is essential to extract the best understanding of the data and making informed SAR decisions.
We have developed a bell-shaped biphasic model to allow double sigmoidal curve fitting, with parameters to describe both sigmoidal parts of the data and crucially, confidence levels around the efficacy. This algorithm overcomes the challenge of achieving convergence unlike complex non-linear statistical models, making it usable in high-throughput settings ( >100 compounds/run). The advanced algorithm we have developed can thus achieve more accurate estimation of IC50 and efficacy to drive medicinal chemistry SAR. We have implemented the fit method in Genedata Screener for the analysis of high throughput experimental data. This method also features automated selection of the correct model, Hill or bell-shaped, and automatic data masking, as appropriate. Application of this method to real world PROTACs data has demonstrated its ability to reduce manual intervention and deliver accurate curve fit parameters, leading to greater confidence in drug SAR insights. Further, the novel curve fit algorithm is able to quantitatively describe data from non-PROTACs targets too, showcasing its flexibility and applicability across projects independent of the underlying biology.