PhD student in the @HDR_UK-@turinginst @wellcometrust PhD Programme in Health Data Science, based at @Cambridge_CL. Interested in Machine Learning in Health.
Excited to share SynBa, my latest work with @carlhenrikek, @MagnusRattray and @marta_milo. SynBa is a new method for the estimation of drug combination synergies with uncertainty quantification.
Preprint link: biorxiv.org/content/10.1101/…
Furthermore, SynBa provides improved accuracy of dose-response predictions and better-calibrated uncertainty estimation for the parameters and the predictions. By following a principled Bayesian workflow, we do not need to trade off predictive accuracy for uncertainty estimation.
We are particularly excited about what comes after this paper. Ultimately, the goal is that the model will be actionable in real-world scenarios. Stay tuned :)
SynBa is developed with the aim of providing uncertainty estimates for the model outputs of interest, so that actionable decision-making criteria can be derived from the model outputs.
For example, in monotherapy, we may be interested in the potency (e.g. IC50) and the efficacy (e.g. Einf) of a single agent. Figure 3 is an example that shows how the uncertainty estimates in SynBa evolve as more data become available.
Thank you so much Marta for hosting us throughout the 3 days! Also a massive thanks to Vera and Chris @cwcyau for making it happen! It has been a brilliant and unique experience for all of us.